id
int64 393k
2.82B
| repo
stringclasses 68
values | title
stringlengths 1
936
| body
stringlengths 0
256k
⌀ | labels
stringlengths 2
508
| priority
stringclasses 3
values | severity
stringclasses 3
values |
|---|---|---|---|---|---|---|
97,423,291 | opencv | GPU HOG bug with non-standard window size | Transferred from http://code.opencv.org/issues/2874
```
|| Thorsten Franzel on 2013-03-08 15:06
|| Priority: Normal
|| Affected: 2.4.4
|| Category: gpu (cuda)
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## GPU HOG bug with non-standard window size
```
Hello,
i suspect i found a offset bug of some sort in the GPU HOG code. I compared to the cpu version.
* When 64x128 (std pedestrian) is used, everything is fine. (cpu and gpu ok)
* When non-standard window sizes are used in the gpu version, something goes wrong. partially really big number like 2.7*10^(+-13)
i transfered and visualized the effect with matlab, because im not yet so familiar with plotting in opencv.
sample code has only average difference which increases as you go away from std pedestrian window size.
Kind regards,
Thorsten
```
## History
##### Thorsten Franzel on 2013-03-11 13:11
```
error occures also on the most recent git revision
is any additional information needed? The CUDA device is a GeForce GTX 285
```
##### Alexander Smorkalov on 2013-03-12 05:41
```
- Assignee set to Marina Kolpakova
- Category set to gpu (cuda)
```
##### Marina Kolpakova on 2013-04-05 23:29
```
- Assignee deleted (Marina Kolpakova)
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: gpu/cuda (contrib) | low | Critical |
97,423,329 | opencv | Stream is blocked by 20 seconds or more when reading an mjpeg stream at low resolutions | Transferred from http://code.opencv.org/issues/2877
```
|| Dustin Spicuzza on 2013-03-10 06:56
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.5
|| Category: highgui-video
|| Tracker: Bug
|| Difficulty: Easy
|| PR:
|| Platform: Any / Any
```
## Stream is blocked by 20 seconds or more when reading an mjpeg stream at low resolutions
```
When opening a network-based mjpeg stream using OpenCV's VideoCapture, the stream doesn't actually get displayed until 20 or 30 seconds later, and the program blocks waiting for ffmpeg to evaluate the stream parameters (as it freezes in avformat_find_stream_info). I find that I don't notice it with 640x480 @ 30fps streams, but at 320x260 @ 5fps the delay is huge. If one enables the ffmpeg logging, after the freeze is done you can see a message similar to "max_analyze_duration 5000000 reached at 5000000".
The following python code can reproduce the problem.
<pre>
import cv2
vc = cv2.VideoCapture()
vc.open('http://192.168.0.10/mjpg/video.mjpg') # freezes on this line
</pre>
I found that you can fix this problem by inserting "ic->max_analyze_duration = 50000;" at line 532 (in CvCapture_FFMPEG::open(), right before avformat_find_stream_info() is called). However, it's not clear to me whether this is a problem that OpenCV should have a setting to fix, or whether this is an ffmpeg problem (though preferably it should be fixed in both locations). It's also not clear to me whether setting the max_analyze_duration affects anything -- as far as I can see, the video stream still works perfectly fine for me.
This affects OpenCV 2.4.4 on Windows (32 bit) and Linux (64 bit). I'm using the Python bindings, but that shouldn't matter at all, I imagine it affects the C++ bindings also.
```
## History
##### Dustin Spicuzza on 2013-03-10 06:59
```
Note that the stream is coming from an AXIS M1011 camera.
```
##### Dustin Spicuzza on 2013-03-10 07:24
```
I filed a bug at FFMpeg also: http://ffmpeg.org/trac/ffmpeg/ticket/2343
```
##### Alexander Smorkalov on 2013-03-11 08:13
```
Good Job!
Thank you for your bug report. OpenCV is open source project and you can contribute your changes via pull request on Github. It will be really faster. Find instruction "here":http://code.opencv.org/projects/opencv/wiki/How_to_contribute. Your change just fixes a bug and does not break compatibility with previous version, so I recommend you to make pull request to branch 2.4.
```
##### Dustin Spicuzza on 2013-03-14 14:28
```
I might do that, but it's not clear to me whether this will break other things that actually use the analyze duration period. For even faster startup time, I set the period to zero, and this seems to work just fine for live video streams (which don't have a duration anyways). I don't have any other video files around to test this on, however, and it's not clear to me what kinds of bugs it could provoke.
```
##### Andrey Kamaev on 2013-04-03 20:06
```
- Target version changed from 2.4.5 to Next Hackathon
```
##### Daniel Angelov on 2013-08-06 22:10
```
I can confirm that this problem(delay) still exists in 2.4.6.1 using the C++ bindings. The delay is inverse proportional to the size of the image - 1600x1200 is around 4 sec.
```
##### Victor Kocheganov on 2013-08-08 06:50
```
- Affected version changed from 2.4.4 to 2.4.0 - 2.4.5
- Operating System set to Any
- HW Platform set to Any
- Assignee changed from Vadim Pisarevsky to Alexander Smorkalov
```
##### Alexander Smorkalov on 2013-08-29 13:31
```
Hello Dustin!
I investigate the issue and solution you proposed. FFMPEG uses this period of time to wait for camera initialization and to analyze stream parameters. You fix decrease this time and can break compatibility with other cameras that are very slow. The only solution applicable here is custom VideoCapture parameter for ffmpeg+ip camera case. Alternatively, you can change encoding format from mjpeg to something else. It decrease initialization time in most cases.
- Difficulty set to Easy
```
##### Dustin Spicuzza on 2013-12-03 04:12
```
Hey Alexander,
The last time I looked, there wasn't a custom parameter available to tune this setting. It would be great if we could add one.
Dustin
```
| wontfix,auto-transferred,category: videoio,affected: 2.4 | low | Critical |
97,423,556 | opencv | [Python] cv2.waitKey returns | Transferred from http://code.opencv.org/issues/2910
```
|| Matthäus Brandl on 2013-03-22 15:18
|| Priority: Normal
|| Affected: 2.4.2
|| Category: highgui-gui
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## [Python] cv2.waitKey returns
```
On Ubuntu 12.10 using either the shipped OpenCV 2.3.1 or OpenCV 2.4.2 of the upcoming Ubuntu 13.04, cv2.waitKey does not return the keycode but 0x100000 + the key code:
<pre>
>>> import cv2
>>> import numpy as np
>>> cv2.imshow("test", np.zeros((10,10)))
>>> key = cv2.waitKey(-1)
>>> print key, hex(key), key % 256
1048603 0x10001b 27
</pre>
(except if the delay runs out, then the result is -1)
See as well this question on StackOverflow: http://stackoverflow.com/questions/9172170
Maybe this bug is related? http://code.opencv.org/issues/191
<pre>
% LC_ALL=C apt-cache policy python-opencv
python-opencv:
Installed: 2.4.2+dfsg-0exp2ubuntu1
Candidate: 2.4.2+dfsg-0exp2ubuntu1
Version table:
*** 2.4.2+dfsg-0exp2ubuntu1 0
100 /var/lib/dpkg/status
2.3.1-11ubuntu2 0
500 http://de.archive.ubuntu.com/ubuntu/ quantal/universe amd64 Packages
</pre>
```
## History
| bug,auto-transferred,priority: normal,category: highgui-gui,affected: 2.4 | low | Critical |
97,423,596 | opencv | [Python] cv2.destroyWindow does not work | Transferred from http://code.opencv.org/issues/2911
```
|| Matthäus Brandl on 2013-03-22 16:04
|| Priority: Normal
|| Affected: 2.4.2
|| Category: highgui-gui
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## [Python] cv2.destroyWindow does not work
```
On Ubuntu 12.10 using either the shipped OpenCV 2.3.1 or OpenCV 2.4.2 of the upcoming Ubuntu 13.04, one cannot close named windows as expected using cv2.destroyWindow.
See for example on StackOverflow:
* http://stackoverflow.com/questions/6116564
* http://stackoverflow.com/questions/11593046
Although both questions use OpenCV 1.x, the same problem occurs using the cv2 import of OpenCV 2.x
The only way to really close a window is to call the functions in this particular order:
<pre>
cv2.destroyWindow(name)
cv2.waitKey(-1)
cv2.imshow(name, img)
</pre>
where img is an arbitrary numpy array.
For the sake of completeness:
<pre>
% LC_ALL=C apt-cache policy python-opencv
python-opencv:
Installed: 2.4.2+dfsg-0exp2ubuntu1
Candidate: 2.4.2+dfsg-0exp2ubuntu1
Version table:
*** 2.4.2+dfsg-0exp2ubuntu1 0
100 /var/lib/dpkg/status
2.3.1-11ubuntu2 0
500 http://de.archive.ubuntu.com/ubuntu/ quantal/universe amd64 Packages
</pre>
```
## History
| bug,auto-transferred,priority: normal,category: highgui-gui,affected: 2.4 | low | Critical |
97,423,633 | opencv | read large .avi file | Transferred from http://code.opencv.org/issues/2913
```
|| Minh Vo on 2013-03-24 06:17
|| Priority: Normal
|| Affected: branch '2.4'
|| Category: highgui-video
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## read large .avi file
```
VideoCapture does not work on Windows. If I disabled FFMPEG when using cmake and recompiled from the source, it partially worked. By partially, I meant it succeeded in reading a number of frames but stop reading after that.
The source can be as simple as this. Run this on a large video file and it will fail!
string fname = "xxx.avi";
VideoCapture cap(fname ); // open the default camera
if(!cap.isOpened()) // check if we succeeded
return -1;
Mat edges;
namedWindow("edges",1);
for(;;)
{
Mat frame;
cap >> frame; // get a new frame from camera
imshow("frame", frame);
if(waitKey(30) >= 0) break;
}
// the camera will be deinitialized automatically in VideoCapture destructor
return 0;
```
## History
##### Daniil Osokin on 2013-03-25 06:37
```
Hi, Minh! Please, attach a link to video file to reproduce it.
- Assignee set to Minh Vo
- Category set to highgui-video
```
| bug,auto-transferred,priority: normal,category: videoio,affected: 2.4 | low | Critical |
97,423,671 | opencv | memory leak in cv:VideoWriter::write() ( iOS ) | Transferred from http://code.opencv.org/issues/2918
```
|| Alessandro Trebbi on 2013-03-25 22:36
|| Priority: Normal
|| Affected: 2.4.4
|| Category: highgui-video
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## memory leak in cv:VideoWriter::write() ( iOS )
```
this function loses memory each time is called ; after some frames , the app crashes
```
## History
##### Alessandro Trebbi on 2013-03-25 22:50
```
Alessandro Trebbi wrote:
> this function loses memory each time is called ; after some frames , the app crashes
I think the bug is here:
cap_avfoundation.mm line 1295
CFDataRef cfData = CGDataProviderCopyData(CGImageGetDataProvider(cgImage));
cfData should be released after use
```
##### Vladislav Vinogradov on 2013-04-01 08:53
```
Thanks for your report.
Could you provide a fix for this issue? See [[How_to_contribute#Fork-38-Pull-Request-model-for-code-contribution]] page for instructions.
- Assignee set to Vadim Pisarevsky
- Category set to highgui-video
```
##### Alessandro Trebbi on 2013-04-01 21:10
```
I did not manage to test it, but this should work:
cap_avfoundation.mm line 1317 i added 2 lines ( i added to release, i'm not sure that both are required )
//cleanup
* NSLog(@"Patch alessandro trebbi");
CFRelease(cfData);
CVPixelBufferRelease(pixelBuffer);*
CGImageRelease(cgImage);
CGDataProviderRelease(provider);
CGColorSpaceRelease(colorSpace);
( i tried to fix it by myself, but did not manage to compile a functional framework, i mean the framework i comoile give me many link errors due to duplicate symbols )
```
##### Alessandro Trebbi on 2013-04-09 13:40
```
Please assign the bug to me, i have already fixed, and then will do a pull request. thx
```
##### Nguyen Nguyen on 2015-05-04 17:45
```
Alessandro Trebbi wrote:
> I did not manage to test it, but this should work:
>
> cap_avfoundation.mm line 1317 i added 2 lines ( i added to release, i'm not sure that both are required )
>
> //cleanup
> * NSLog(@"Patch alessandro trebbi");
> CFRelease(cfData);
> CVPixelBufferRelease(pixelBuffer);*
> CGImageRelease(cgImage);
> CGDataProviderRelease(provider);
> CGColorSpaceRelease(colorSpace);
>
> ( i tried to fix it by myself, but did not manage to compile a functional framework, i mean the framework i comoile give me many link errors due to duplicate symbols )
I tried it on the latest 2.4 branch, the leak is still there.
```
| bug,auto-transferred,priority: normal,category: videoio,affected: 2.4 | low | Critical |
97,423,712 | opencv | createLBPHFaceRecognizer function in Java | Transferred from http://code.opencv.org/issues/2919
```
|| Anton Anikin on 2013-03-26 07:27
|| Priority: Normal
|| Affected: None
|| Category: java bindings
|| Tracker: Feature
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## createLBPHFaceRecognizer function in Java
```
Following functions are not mapped in java to native objects createEigenFaceRecognizer(), createFisherFaceRecognizer(), createLBPHFaceRecognizer()
I did it in following way, maybe it will be useful. I added few rows into modules/java/generator/gen_java.py file:
<pre>
'Contrib' :
{
"createLBPHFaceRecognizer" : {'j_code' :
"""
public static FaceRecognizer createLBPHFaceRecognizer(){
return new FaceRecognizer(n_createLBPHFaceRecognizer());
}
"""
, 'jn_code' :
"""
private static native long n_createLBPHFaceRecognizer();\n
"""
, 'cpp_code' :
"""
JNIEXPORT jlong JNICALL Java_org_opencv_contrib_Contrib_n_1createLBPHFaceRecognizer (JNIEnv *, jclass);
JNIEXPORT jlong JNICALL Java_org_opencv_contrib_Contrib_n_1createLBPHFaceRecognizer (JNIEnv * jenv, jclass) {
cv::Ptr<FaceRecognizer> model = cv::createLBPHFaceRecognizer();
model.addref();
return (jlong)model.obj;
}
"""
}, # createLBPHFaceRecognizer
}, # Contrib
</pre>
```
## History
##### Anton Anikin on 2013-03-26 07:30
```
<pre>
'Contrib' :
{
"createLBPHFaceRecognizer" : {'j_code' :
"""
public static FaceRecognizer createLBPHFaceRecognizer(){
return new FaceRecognizer(n_createLBPHFaceRecognizer());
}
"""
, 'jn_code' :
"""
private static native long n_createLBPHFaceRecognizer();\n
"""
, 'cpp_code' :
"""
JNIEXPORT jlong JNICALL Java_org_opencv_contrib_Contrib_n_1createLBPHFaceRecognizer (JNIEnv *, jclass);
JNIEXPORT jlong JNICALL Java_org_opencv_contrib_Contrib_n_1createLBPHFaceRecognizer (JNIEnv * jenv, jclass) {
cv::Ptr<FaceRecognizer> model = cv::createLBPHFaceRecognizer();
model.addref();
return (jlong)model.obj;
}
"""
}, # createLBPHFaceRecognizer
}, # Contrib
</pre>
```
##### Vladislav Vinogradov on 2013-04-01 09:11
```
- Assignee set to Andrey Pavlenko
- Category set to java bindings
```
##### Koray Gecici on 2013-04-24 06:58
```
We need createLBPHFaceRecognizer method from Java. Can i download from anywhere, including full version of Contrib java class?
```
##### Anna Kogan on 2013-05-07 09:19
```
- Description changed from Following functions are not mapped in java
to native objects createEigenFaceR... to Following functions are not
mapped in java to native objects createEigenFaceR... More
```
##### Jay ZH on 2013-08-09 23:05
```
This is not correct.
createLBPHFaceRecognizer defined as
Ptr<FaceRecognizer> createLBPHFaceRecognizer(int radius, int neighbors,
int grid_x, int grid_y, double threshold)
need a java binding match native definition.
```
| auto-transferred,priority: normal,feature,category: java bindings | low | Minor |
97,423,779 | opencv | Discrete wavelet transform | Transferred from http://code.opencv.org/issues/2923
```
|| Ajay Viswanathan on 2013-03-26 20:48
|| Priority: Normal
|| Affected: None
|| Category: imgproc, video
|| Tracker: Feature
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## Discrete wavelet transform
```
There should functions for Haar, DB-X series, 9/7 and all other kinds of transforms in the OpenCV libraries. They would be quite handy in image compression and transmission applications.
```
## History
##### Vladislav Vinogradov on 2013-04-01 09:06
```
Hello, Ajay Viswanathan.
OpenCV is open source library.
You can make your contribution to it. If you can implement this feature, submit a pull request (see [[How_to_contribute]]).
```
##### Rel Guzman Apaza on 2013-04-02 18:28
```
Vladislav Vinogradov wrote:
> Hello, Ajay Viswanathan.
>
> OpenCV is open source library.
> You can make your contribution to it. If you can implement this feature, submit a pull request (see [[How_to_contribute]]).
How large should be the code?
https://github.com/Itseez/opencv/pull/4#issuecomment-7393721
```
##### Rel Guzman Apaza on 2013-04-02 18:28
```
Vladislav Vinogradov wrote:
> Hello, Ajay Viswanathan.
>
> OpenCV is open source library.
> You can make your contribution to it. If you can implement this feature, submit a pull request (see [[How_to_contribute]]).
How large should be the code?
[[https://github.com/Itseez/opencv/pull/4#issuecomment-7393721]]
```
##### Vladislav Vinogradov on 2013-04-03 05:44
```
First of all, new functionality must be useful for large number of applications.
```
##### Vladislav Vinogradov on 2013-04-03 05:45
```
- Category set to imgproc, video
```
| auto-transferred,feature,category: imgproc | low | Minor |
97,423,804 | opencv | Need to update Introduction to Java Development: java sun 6 on Ubuntu 12.04 | Transferred from http://code.opencv.org/issues/2935
```
|| Oleg Sklyarov on 2013-04-02 07:55
|| Priority: Low
|| Affected: None
|| Category: documentation
|| Tracker: Feature
|| Difficulty: Easy
|| PR:
|| Platform: None / None
```
## Need to update Introduction to Java Development: java sun 6 on Ubuntu 12.04
```
There are two alternative solutions:
# https://github.com/flexiondotorg/oab-java6
# http://source.android.com/source/initializing.html#installing-the-jdk
```
## History
##### Kirill Kornyakov on 2013-07-08 07:53
```
- Target version changed from 2.4.6 to Next Hackathon
- Difficulty set to Easy
- Priority changed from Normal to Low
```
| auto-transferred,feature,priority: low,category: documentation | low | Minor |
97,423,921 | opencv | Qt Window: Save image button does not work (Bug #2449) | Transferred from http://code.opencv.org/issues/2947
```
|| Stefan R on 2013-04-09 12:11
|| Priority: Normal
|| Affected: branch 'master' (2.4.9)
|| Category: highgui-gui
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## Qt Window: Save image button does not work (Bug #2449)
```
Please re-open the ticket "Qt Window: Save image button does not work (Bug #2449)".
It is still not fixed. See comments in Bug #2449.
```
## History
##### Maria Dimashova on 2013-04-09 16:34
```
Can you please specify the OpenCV version or revision you used as well as your environment? It would even better if you propose a pull request that fixes and tests the problem. Thanks.
- Target version deleted (2.4.6)
```
##### Stefan R on 2013-04-10 07:01
```
It is all mentioned in #2449. That's why I refer to this particular ticket.
This bug is present in both Git branches "2.4" and "master"!
Git version as of Friday 5.4.2013, Windows 7 x64, OpenCV built with Qt 4.8.4.
And by the way: The patch of Cenk Uyan does fix it if the Qt windows displays still images (e.g. with cv::waitKey(0)) but not for videos (e.g with cv::waitkey(10)).
```
| bug,auto-transferred,priority: normal,category: highgui-gui,affected: 2.4 | low | Critical |
97,423,955 | opencv | PNG imwrite with ICC | Transferred from http://code.opencv.org/issues/2948
```
|| Florian Meier on 2013-04-09 13:53
|| Priority: Normal
|| Affected: None
|| Category: highgui-images
|| Tracker: Feature
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## PNG imwrite with ICC
```
I would like to store a color profile together with the PNG files.
For this, I modified the grfmt_png.cpp and added the lines
@if(profileData != NULL)
png_set_iCCP(png_ptr, info_ptr, "ICC Profile", PNG_COMPRESSION_TYPE_BASE, profileData, profileDataLength);
@
within the write function and this works very well.
I think this could also be a handy feature for other users, but I don't know how the API for setting the profileData should look like, because the write function is normally not directly called by the user. Furthermore, using the params vector could be working, but would require ugly pointer casting.
Any ideas?
Greetings,
Florian
```
## History
##### Maria Dimashova on 2013-04-09 16:07
```
Florian, could you please create a pull request with your suggestion on "OpenCV github":https://github.com/Itseez/opencv/pulls?
- Assignee set to Vadim Pisarevsky
- Category set to highgui-images
```
| wontfix,auto-transferred,category: imgcodecs | low | Minor |
97,423,998 | opencv | Conversion RGB2HSI | Transferred from http://code.opencv.org/issues/2950
```
|| Rel Guzman Apaza on 2013-04-10 00:16
|| Priority: Normal
|| Affected: None
|| Category: imgproc, video
|| Tracker: Feature
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## Conversion RGB2HSI
```
There is no CV_RGB2HSI conversion in the github repository
I need the function to implement something like this one
http://www.sciencedirect.com/science/article/pii/S1077314206002141
I can do the changes in the repository .')
```
## History
##### Maria Dimashova on 2013-04-10 07:18
```
Yes, please do the changes and submit them as a pull request on "OpenCV github":https://github.com/Itseez/opencv/pulls.
- Priority changed from High to Low
- Category set to imgproc, video
```
##### Maria Dimashova on 2013-04-10 07:19
```
- Priority changed from Low to Normal
```
##### Kirill Kornyakov on 2013-04-11 14:31
```
- Assignee changed from Vadim Pisarevsky to Rel Guzman Apaza
```
##### Gurpinder Sandhu on 2013-04-26 16:40
```
Any updates whether it has been done, I want to use it. If not I myself can make a pull request :)
```
##### Anna Kogan on 2013-05-07 08:40
```
Hello Rel, Gurpinder,
You are welcome to make pull request implementing the functionality. Here's full description of contributing process: http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute
```
| auto-transferred,priority: normal,feature,category: imgproc,category: video | low | Minor |
97,424,062 | opencv | videostab: missing documentation | Transferred from http://code.opencv.org/issues/2955
```
|| Tom Deakin on 2013-04-14 14:22
|| Priority: Low
|| Affected: 2.4.5 (latest release)
|| Category: documentation
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## videostab: missing documentation
```
In the current http://docs.opencv.org/ there is no documentation for the videostab module.
Also, the documentation at http://docs.opencv.org/trunk/modules/videostab/doc/videostab.html is incomplete - there is no information about the GaussianMotionFilter (for example) and others used in the videostab.cpp sample file.
As videostab is included in the current 2.4.5 release without documentation, would it be possible to generate the documentation for this?
```
## History
##### Kirill Kornyakov on 2013-04-14 20:54
```
Tom, it would be great if you could work on it yourself and prepare a pull request. Reading code should be simpler than writing =) Original author of this module has left the project, so better to not expect fast response.
Is anybody else from community ready to work on documenting this?
- Target version set to Next Hackathon
- Priority changed from High to Low
- Category changed from videostab to documentation
```
##### Tom Deakin on 2013-04-14 23:23
```
Kirill, Thanks for the update. I will see if I have any time and will update some documentation. Just to clarify, the 2.4 branch is missing some of the updates from master, some of which were written in 2012. Do you know why these changed have not been included in a 2.4.x release yet?
TLDR: which branch is the documentation for?
```
| bug,auto-transferred,priority: low,category: documentation,affected: 2.4 | low | Critical |
97,424,091 | opencv | Flann problem with AutotunedIndexParams | Transferred from http://code.opencv.org/issues/2956
```
|| Fisher Yu on 2013-04-14 15:35
|| Priority: Normal
|| Affected: branch '2.4'
|| Category: flann
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## Flann problem with AutotunedIndexParams
```
I use flann with AutotunedIndexParams to match sift descriptors. However, when there were more than 1000 features, I got the error "pointer being freed was not allocated". It did work for several hundred features and I don't know the exact boundary number of features to cause the problem. Attached is a small program to reproduce the problem. I am using 2.4.3 with clang on Mac OS X 10.8
```
## History
##### Vladislav Vinogradov on 2013-06-10 08:19
```
- Category set to flann
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: flann | low | Critical |
97,424,120 | opencv | Multiple vulnerabilities in OpenCV | Transferred from http://code.opencv.org/issues/2968
```
|| Andres Gomez on 2013-04-17 04:02
|| Priority: Low
|| Affected: None
|| Category: core
|| Tracker: Feature
|| Difficulty:
|| PR:
|| Platform: None / None
```
## Multiple vulnerabilities in OpenCV
```
Hi,
I worked with OpenCv some time ago and I decided to carry out a little security audit. The tool used was flawfinder with the following command line:
flawfinder -m 3 --html --quiet --dataonly --context opencv-2.4.4/ > opencv-flaws.html
This analyses the source code and returns a lot of results, most of them false positive. But there are already several vulnerabilities I checked as real:
In the file opencv-2.4.4/modules/legacy/src/oneway.cpp line 1795:
if (fscanf(pFile, "%s", imagename) <= 0)
There is a buffer overflow when reading long files names from a list of files. For example when executing:
./c-example-one_way_sample location scene_l.bmp scene_r.bmp
Reading the images...
Extracted 42 keypoints...
Training one way descriptors...
*** buffer overflow detected ***: ./c-example-one_way_sample terminated
This program reads a file called "one_way_train_images.txt" which its content is:
one_way_train_0000AAAAAAAAAAA...AAAAAAAAAAAAAAAAAAAAAAAAAA.jpg
one_way_train_0001BBBBBBBBBBB...BBBBBBBBBBBBBBBBBBBBBBBBBB.jpg
In the file opencv-2.4.4/modules/highgui/src/cap_images.cpp line 114:
sprintf(str, filename, firstframe + currentframe);
There is a format string when reading special files names. For example:
./c-example-adaptiveskindetector C:\VideoSequences\sample1\right_view\temp_%05d%n%s%s%s%s%s.jpg 0 1000
Press ESC to stop.
*** %n in writable segment detected ***
Abortado
These bugs can be exploited to execute arbitrary code.
Although of less concern, also there are vulnerabilities in the samples, here are two I found:
opencv-2.4.4/samples/cpp/hybridtrackingsample.cpp line 82, a buffer overflow:
sprintf(test_file, "%s", argv[1]);
opencv-2.4.4/samples/cpp/hybridtrackingsample.cpp line 85 another buffer overflow:
int values_read = fscanf(f, "%s\n", vid);
Best regards,
Andres Gomez.
```
## History
##### Vadim Pisarevsky on 2015-04-28 11:49
```
thanks; I do not think we will be able to address any of the vulnerabilities by ourselves, we are not experts in it. We'd let specialists to provide patches that we will be happy to integrate
- Priority changed from Normal to Low
- Tracker changed from Bug to Feature
```
| auto-transferred,feature,priority: low,category: core | low | Critical |
97,424,253 | opencv | SIFT Octave bug | Transferred from http://code.opencv.org/issues/2987
```
|| Andew Sabrew on 2013-04-25 20:36
|| Priority: Normal
|| Affected: None
|| Category: nonfree
|| Tracker: Feature
|| Difficulty:
|| PR:
|| Platform: None / None
```
## SIFT Octave bug
```
I think SIFT sets a bad value for octave in the @KeyPoint@ class. If you print the value of keypoint.octave you get something like this:
<pre>
Octave value:10420482
Octave value:10420482
Octave value:11337985
Octave value:10944771
</pre>
And it can't be true because octave ranges are very low.
```
## History
##### Anna Kogan on 2013-05-07 08:48
```
Hello Andrew,
Thank you for reporting the issue. If you investigate the problem further and fix it on your side, a "pull request":http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute in our GitHub repo would be highly appreciated!
```
##### Vladislav Vinogradov on 2013-06-10 10:35
```
SIFT writes a packed value (octave, layer and scale) to octave member. You can unpack this values with following function:
<pre>
void unpackSIFTOctave(const KeyPoint& kpt, int& octave, int& layer, float& scale)
{
octave = kpt.octave & 255;
layer = (kpt.octave >> 8) & 255;
octave = octave < 128 ? octave : (-128 | octave);
scale = octave >= 0 ? 1.f/(1 << octave) : (float)(1 << -octave);
}
</pre>
- Status changed from Open to Cancelled
```
##### Guanta Alecho on 2013-06-21 21:30
```
Sorry to start a discussion: But shouldn't this be changed? Hhow should a typical user know that?
- Status changed from Cancelled to Incomplete
```
##### Kirill Kornyakov on 2013-06-27 08:31
```
Well, we can either document this, or change the behavior. Most likely we shouldn't change the behavior, so the existing code continue to work. Where it could be done? Or you have other suggestions?
- Description changed from I think SIFT sets a bad value for octave in
the KeyPoint class. If you print ... to I think SIFT sets a bad
value for octave in the @KeyPoint@ class. If you prin... More
- Category set to nonfree
- Tracker changed from Bug to Feature
```
| auto-transferred,priority: normal,feature,category: nonfree | low | Critical |
97,424,353 | opencv | opencv_traincascade crashing with maxDepth > 3 | Transferred from http://code.opencv.org/issues/2991
```
|| Arvid Terzibaschian on 2013-04-26 14:52
|| Priority: Normal
|| Affected: 2.4.5 (latest release)
|| Category: objdetect
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## opencv_traincascade crashing with maxDepth > 3
```
I used opencv_traincascade with a clean clone from git opencv master, as well as 2.4.5.
If you set the maxDepth parameter > 3 opencv_traincascade segfaults when writing the trained stage.
I was able to hack-fix it and remove training, but someone with a deeper knowlegde of the code should find a cleaner fix.
Using gdb and valgrind I found 3 bugs related to the problem which are:
1)
<pre>
==3024== Conditional jump or move depends on uninitialised value(s)
==3024== at 0x528CD9F: CvBoostTree::find_split_cat_class(CvDTreeNode*, int, float, CvDTreeSplit*, unsigned char*) (boost.cpp:410)
==3024== by 0x52766D3: cv::DTreeBestSplitFinder::operator()(cv::BlockedRange const&) (tree.cpp:1918)
==3024== by 0x527680E: CvDTree::find_best_split(CvDTreeNode*) (internal.hpp:251)
==3024== by 0x526FACF: CvDTree::try_split_node(CvDTreeNode*) (tree.cpp:1699)
==3024== by 0x528AAB5: CvBoostTree::try_split_node(CvDTreeNode*) (boost.cpp:163)
==3024== by 0x526F02F: CvDTree::do_train(CvMat const*) (tree.cpp:1645)
==3024== by 0x41C9A5: CvCascadeBoost::train(CvFeatureEvaluator const*, int, int, int, CvCascadeBoostParams const&) (boost.cpp:1302)
==3024== by 0x41318A: CvCascadeClassifier::train(std::string, std::string, std::string, int, int, int, int, int, CvCascadeParams const&, CvFeatureParams const&, CvCascadeBoostParams const&, bool) (cascadeclassifier.cpp:213)
==3024== by 0x40BD84: main (traincascade.cpp:110)
==3024==
</pre>
This is just a minor issue caused by split_criteria not being initialized properly. I fixed it by
<pre>
CvBoostParams::CvBoostParams()
{
split_criteria = CvBoost::DEFAULT;
boost_type = CvBoost::REAL;
weak_count = 100;
weight_trim_rate = 0.95;
cv_folds = 0;
max_depth = 1;
}
</pre>
2) valgrind furthermore reported several invalid writes during tree-export all looking like this:
<pre>
==3024== Invalid write of size 4
==3024== at 0x41D981: CvCascadeBoostTree::write(cv::FileStorage&, cv::Mat const&) (boost.cpp:927)
==3024== by 0x41ED11: CvCascadeBoost::write(cv::FileStorage&, cv::Mat const&) const (boost.cpp:1645)
==3024== by 0x413624: CvCascadeClassifier::train(std::string, std::string, std::string, int, int, int, int, int, CvCascadeParams const&, CvFeatureParams const&, CvCascadeBoostParams const&, bool) (cascadeclassifier.cpp:248)
==3024== by 0x40BD84: main (traincascade.cpp:110)
==3024== Address 0xa279538 is 0 bytes after a block of size 8 alloc'd
==3024== at 0x4C225BA: operator new[](unsigned long) (vg_replace_malloc.c:264)
==3024== by 0x41D1C2: CvCascadeBoostTree::write(cv::FileStorage&, cv::Mat const&) (boost.cpp:911)
==3024== by 0x41ED11: CvCascadeBoost::write(cv::FileStorage&, cv::Mat const&) const (boost.cpp:1645)
==3024== by 0x413624: CvCascadeClassifier::train(std::string, std::string, std::string, int, int, int, int, int, CvCascadeParams const&, CvFeatureParams const&, CvCascadeBoostParams const&, bool) (cascadeclassifier.cpp:248)
==3024== by 0x40BD84: main (traincascade.cpp:110)
</pre>
This is caused by a wrong calculation of the number of to be expected tree nodes in function
<pre>
void CvCascadeBoostTree::write( FileStorage &fs, const Mat& featureMap )
</pre>
When I changed
<pre>
int size = (int)pow( 2.f, (float)ensemble->get_params().max_depth);
Ptr<float> leafVals = new float[size];
</pre>
to
<pre>
int size = (int)pow( 2.f, (float)ensemble->get_params().max_depth) * 50;
Ptr<float> leafVals = new float[size];
</pre>
it worked again - valgrind reported no more illegal writes.
This is obviously a bad hack and should be fixed by someone who better knows how to calculate this value correctly.
3) valgrind reported another issue:
<pre>
==4327== Mismatched free() / delete / delete []
==4327== at 0x4C21A33: operator delete(void*) (vg_replace_malloc.c:346)
==4327== by 0x41A656: CvCascadeBoostTree::write(cv::FileStorage&, cv::Mat const&) (cvstd.hpp:392)
==4327== by 0x41AB4F: CvCascadeBoost::write(cv::FileStorage&, cv::Mat const&) const (boost.cpp:1697)
==4327== by 0x413D08: CvCascadeClassifier::train(std::string, std::string, std::string, int, int, int, int, int, CvCascadeParams const&, CvFeatureParams const&, CvCascadeBoostParams const&, bool) (cascadeclassifier.cpp:247)
==4327== by 0x40CA54: main (traincascade.cpp:109)
==4327== Address 0xa4dbbb0 is 0 bytes inside a block of size 400 alloc'd
==4327== at 0x4C225BA: operator new[](unsigned long) (vg_replace_malloc.c:264)
==4327== by 0x419A4C: CvCascadeBoostTree::write(cv::FileStorage&, cv::Mat const&) (boost.cpp:960)
==4327== by 0x41AB4F: CvCascadeBoost::write(cv::FileStorage&, cv::Mat const&) const (boost.cpp:1697)
==4327== by 0x413D08: CvCascadeClassifier::train(std::string, std::string, std::string, int, int, int, int, int, CvCascadeParams const&, CvFeatureParams const&, CvCascadeBoostParams const&, bool) (cascadeclassifier.cpp:247)
==4327== by 0x40CA54: main (traincascade.cpp:109)
==4327==
</pre>
I did not further investigate it because the problem has been hack-fixed with fix 1) and 2), but that should be fixed aswell.
Hope that helps, seeing forward to a fix soon.
```
## History
##### Shengyin Wu on 2013-04-30 02:04
```
i have try maxDepth parameter > 3, but the program run correctlly.
```
##### Anna Kogan on 2013-05-07 08:29
```
Hello Shengyin,
Thank you for the full description of the problem. If you find solution on your side, please send a "pull request":http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute in our GitHub repo.
- Assignee set to Marina Kolpakova
- Category set to objdetect
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: objdetect | low | Critical |
97,424,395 | opencv | convolveDFT example code in documentation is incorrect | Transferred from http://code.opencv.org/issues/2997
```
|| Alan Gardiner on 2013-04-30 02:45
|| Priority: Normal
|| Affected: 2.4.5 (latest release)
|| Category: core
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## convolveDFT example code in documentation is incorrect
```
The size that is allocated for the resulting matrix in the convolveDFT example code is incorrect and at best misleading. In order to perform a full 2D convolution the size of Mat C should not be
<pre>
void convolveDFT(InputArray A, InputArray B, OutputArray C)
{
// reallocate the output array if needed
C.create(abs(A.rows - B.rows)+1, abs(A.cols - B.cols)+1, A.type());
...
</pre>
as indicated in the dft documentation. This example code only gives the convolution where the images are completely overlapped. In the limiting case where A and B are the same sizes, this code gives a single point, which is not very useful.
A beginning image processing class teaches that the result of convolving images or matrices A and B should be
<pre>
C.create(A.rows + B.rows-1,A.cols + B.cols-1, A.type());
</pre>
```
## History
##### Anna Kogan on 2013-05-07 08:23
```
Hello Alan,
Thank you for reporting the issue. If you send a corresponding "pull request":http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute in our GitHub repo would be highly appreciated!
- Description changed from The size that is allocated for the
resulting matrix in the convolveDFT exampl... to The size that is
allocated for the resulting matrix in the convolveDFT exampl... More
- Assignee set to Alan Gardiner
- Category set to core
```
| bug,auto-transferred,priority: normal,category: core,affected: 2.4 | low | Critical |
97,424,453 | opencv | cv::TriangulatePoints() | Transferred from http://code.opencv.org/issues/2998
```
|| Ander Biguri on 2013-04-30 09:29
|| Priority: Normal
|| Affected: None
|| Category: calibration, 3d
|| Tracker: Feature
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## cv::TriangulatePoints()
```
I am new to this so not sure If doing this correctly.
I think openCV needs the @triangulatePoints()@ function to be developed more. It needs documentation and flags, among different method implementation, such as Linear/iterative LS ,Linear/iterative Eigen, Poly...
```
## History
##### Anna Kogan on 2013-05-07 08:17
```
Hello Ander,
Thank you for reporting the issue. If you could implement such functionality on your side, a "pull request":http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute in our GitHub repo would be highly appreciated!
```
| auto-transferred,priority: normal,feature,category: calib3d | low | Minor |
97,424,542 | opencv | libv4l capturing backend spams stderr with "HIGHGUI ERROR: V4L2:..." | Transferred from http://code.opencv.org/issues/3004
```
|| Hans Meine on 2013-05-03 09:00
|| Priority: Normal
|| Affected: branch 'master' (2.4.9)
|| Category: highgui-camera
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## libv4l capturing backend spams stderr with "HIGHGUI ERROR: V4L2:..."
```
I have tried querying various properties from the capturing backend, but the libv4l backend defeats this by writing errors to stderr every time I try to access unsupported properties.
AFAICS, unsupported properties should be ignored (that is, setProperty returns false, but unfortunately, getProperty cannot do so ATM). If I want to write a backend-independent client software, it is undesirable to get lots of errors on specific backends, no?
Even better would be to have a way to query (un)supported properties, e.g. by passing an optional bool* to getProperty, and let it write a success flag to that pointer if it is non-NULL.
Finally, I wonder whether these "ERROR"s fprintf's should have been CV_WARN()s like e.g. in cap_gstreamer.cpp.
```
## History
##### Hans Meine on 2013-05-03 09:02
```
BTW, I forgot to mention that issue #1984 seems to be slightly related, but I think it is more about a specific instance of this problem.
```
##### Anna Kogan on 2013-05-07 08:01
```
Hello Hans,
Thank you for reporting the issue. If you can solve the issue on your side, a "pull request":http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute in our GitHub repo would be highly appreciated!
- Assignee changed from Vadim Pisarevsky to Hans Meine
```
##### Hans Meine on 2013-05-07 09:28
```
Hi Anna, does your comment imply that my thoughts are acknowledged? I don't want to spend time on this if you don't agree.
```
##### Anna Kogan on 2013-05-07 09:30
```
- Assignee changed from Hans Meine to Vadim Pisarevsky
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: videoio(camera) | low | Critical |
97,424,573 | opencv | Request: Bag of Words from precomputed keypoints and descriptors | Transferred from http://code.opencv.org/issues/3005
```
|| antonio sesto on 2013-05-03 09:35
|| Priority: Low
|| Affected: None
|| Category: features2d
|| Tracker: Feature
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## Request: Bag of Words from precomputed keypoints and descriptors
```
Hello,
the API for implementing an algorithm based on the Bag-of-Words (BoW) approach does not contain anything for computing the BoW representation of an image from a pre-computed vocabulary, a pre-computed set of keypoints, and a pre-computed set of descriptors.
Once the vocabulary has been computed, the only available class for producing a BoW representation of an input image is:
[[http://docs.opencv.org/modules/features2d/doc/object_categorization.html?highlight=descriptorextractor#bowimgdescriptorextractor]]
that requires an extractor and a matcher.
Even if it is not hard to overcome this limitation, it would be a nice addition a class where you only need to specify the vocabulary and the descriptors.
```
## History
##### Anna Kogan on 2013-05-07 07:59
```
Hello Antonio,
Thank you for reporting the issue. If you could fix the problem on your side, a "pull request":http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute in our GitHub repo would be highly appreciated!
- Assignee set to antonio sesto
- Category set to features2d
```
| auto-transferred,feature,priority: low,category: features2d | low | Minor |
97,424,676 | opencv | Counterintuitive comparison of matrices with NaN values | Transferred from http://code.opencv.org/issues/3011
```
|| Leonid Beynenson on 2013-05-06 13:17
|| Priority: Low
|| Affected: branch '2.4'
|| Category: core
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## Counterintuitive comparison of matrices with NaN values
```
At the moment element-wise comparison of matrices works in OpenCV as follows:
* the comparison operations @A < B@ (and @A > B@) are calculated element-wise
* but the comparison operations @A <= B@ (and @A >= B@) are calculated as the logical negation of the compariosn @A > B@ (correspondingly, @A <B@)
(see the code of the function @cv::compare@ in the file source:modules/core/src/arithm.cpp and the template function @cmp_@ in the same file)
Since in C++ the operations @<@, @>@, @<=@, and @>=@ with NaN values always return @false@, there is the following counterintuitive behavior:
if a matrix A contains only NaN values, we will have
<pre>
countNonZero(A < 0.5) ==> 0;
countNonZero(A > 0.5) ==> 0;
countNonZero(A <= 0.5) ==> A.total();
countNonZero(A >= 0.5) ==> A.total();
</pre>
This may be fixed if the comparison functions will make element-wise calculations for non-strict inequalities too. But in this case all the optimizations (SSE, etc) should be changed too.
```
## History
| bug,auto-transferred,priority: low,category: core,affected: 2.4 | low | Critical |
97,424,852 | opencv | Orb detector-Number of detected areas-BAD formulation | Transferred from http://code.opencv.org/issues/3017
```
|| Zoltan Prohaszka on 2013-05-11 15:37
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.4
|| Category: features2d
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## Orb detector-Number of detected areas-BAD formulation
```
The detector phase of the ORB method detects many-many keypoints.
It is far too much for most of the applications, so it contains
an algorithm to keep the few best keypoints.
This part of the algorithm is bad in the 2.4.4 version.
It assumes that a given scale would have a portion of required number
of keypoints proportional to the linear scale of the scale.
The proper would be proportional to the square of the linear scale.
I tested it, the corrected formulation performs much better
in a registration application. I did not remember wether the basic ORB
article is describing such details or not.
The corrected code (original commented):
orb.cpp:
static void computeKeyPoints(const vector<Mat>& imagePyramid,
const vector<Mat>& maskPyramid,
vector<vector<KeyPoint> >& allKeypoints,
int nfeatures, int firstLevel, double scaleFactor,
int edgeThreshold, int patchSize, int scoreType )
{
int nlevels = (int)imagePyramid.size();
vector<int> nfeaturesPerLevel(nlevels);
// fill the extractors and descriptors for the corresponding scales
//calculate desited numbers per scale:
//original:
// float factor = (float)(1.0 / scaleFactor);
// float ndesiredFeaturesPerScale = nfeatures*(1 - factor)/(1 - (float)pow((double)factor, (double)nlevels));
//Modification:
float factor = (float)(1.0 / scaleFactor);
float factor2D=factor*factor;//scalefactor is linear, factor is areal
float ndesiredFeaturesPerScale = nfeatures*(1 - factor2D)/(1 - (float)pow((double)factor2D, (double)nlevels));
// ratio: curr_scale/all correction since the series doesnt sum to infinity
Our projects needs the corrected functionality
under Android too. I do not know when will I be able to
get my GIT working properly, so if anybody would be so kind
to modify the source please do it and give feedback!
Best regards
zol
```
## History
##### Rui Marques on 2013-11-26 12:57
```
This seems like a useful patch, anyone cares to make a pull request?
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: features2d | low | Critical |
97,424,885 | opencv | SVD in Lapack is suboptimal, by means of Givens rotations vs. Hauseholder mirroring | Transferred from http://code.opencv.org/issues/3018
```
|| Zoltan Prohaszka on 2013-05-11 16:17
|| Priority: Normal
|| Affected: None
|| Category: core
|| Tracker: Feature
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## SVD in Lapack is suboptimal, by means of Givens rotations vs. Hauseholder mirroring
```
Most Multiple View Computer Vision Algorithms have to
solve overdetermined homogeneous linear equations, for
which Singular Value Decomposition (SVD) is the proper method.
The SVD implementation given in Lapack achieves band diagonal shape
by applying Givens rotations. Hausholder mirroring transformations
can zero a partial row for once, not only one element, thus there are
faster by a factor of n. (the smallest size of the input matrix)
A typipical 3D reconstruction algorithm may spend many time by finding inliers
of a proper motion stereo model by RANSAC.
For a not so easy image pair, RANSAC (inside findfundamental())
executes 8x8 SVD several hundred times.
An other example, the JAMA package of NIST uses Hauseholder transformations.
(I attach a commented version of the algorithm, and the original
file also). The JAMA package have another attractive feature regarding
the eigenvalue-decomposition of non-symmetric matrices: It contains
the random step introduced by the MATLAB team to avoid the algorithm to stuck.
(It can be tested by big permutation matrices).
The JAMA package comes with very liberate licence.
It has also disadvantages:
-Most numerical loops are designed to work only with rectangular matrices of
one orientation ('landscape' or 'portrait'), so wrapper routines would be needed
for universal functionality.
-It is not implemented to take advantage of SIMD or multithreaded execution.
-It contains some calculations which give numeric stability very near to
the margin of floating numeric represantations, but results slower execution
in other, common cases (hypot and division instead of multiplication by reciprocal).
I did not analize how unsafe the faster versions would be, so I did not modify these
for my versions.
I would like to ask for the opinion of the community, how the better functionality
should be incorporated in openCV.
Regards:
zol
I set the priority normal because SVD is executed frequently. Otherwise I would give Low.
```
## History
| auto-transferred,priority: normal,feature,category: core | low | Major |
97,424,920 | opencv | CvCapture_FFMPEG::retrieveFrame(...) gives SIGSEGV for high-framerate AVI | Transferred from http://code.opencv.org/issues/3030
```
|| Brian TheLion on 2013-05-16 18:48
|| Priority: Normal
|| Affected: 2.4.5 (latest release)
|| Category: highgui-video
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## CvCapture_FFMPEG::retrieveFrame(...) gives SIGSEGV for high-framerate AVI
```
Basic video capture-and-display scenario using compiled 2.4.5 libraries:
<pre>
/* cvvideoshow.cpp */
#include <opencv2/highgui/highgui.hpp>
#include <iostream>
using namespace cv;
using namespace std;
int main( int argc, char** argv ) {
if( argc != 2) {
cout <<" Usage: cvimshow [video_file]" << endl;
return -1;
}
VideoCapture cap(argv[1]);
if (!cap.isOpened()) { //check if video device has been initialized
cout << "Cannot open file" << endl;
}
Mat cameraFrame;
//unconditional loop
while (true) {
if (!cap.read(cameraFrame)) break;
imshow("cam", cameraFrame);
if (waitKey(30) >= 0) break;
}
return 0;
}
</pre>
<pre>
$ gcc cvvideoshow.cpp -o cvvideoshow -lopencv_highgui -lopencv_core -lstdc++
$ ./cvvideoshow 1000fps.avi # file available for download at https://docs.google.com/file/d/0Bz_2ACaA7oKEV2d6eF93WDd6YVk/edit?usp=sharing
Segmentation Fault
$ (install)br@breach:~/dev/proj/cv/build/cvhelpers/src/cvimshow$ gdb --args ./cvimshow 1000fps.avi
GNU gdb (GDB) 7.5.91.20130417-cvs-ubuntu
Copyright (C) 2013 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "x86_64-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
Reading symbols from cvvideoshow...(no debugging symbols found)...done.
(gdb) r
Starting program: ./cvvideoshow 1000fps.avi
warning: no loadable sections found in added symbol-file system-supplied DSO at 0x7ffff7ffa000
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
Program received signal SIGSEGV, Segmentation fault.
0x00007ffff31c3e51 in ?? ()
from /usr/lib/x86_64-linux-gnu/libswscale.so.2
(gdb) bt
#0 0x00007ffff31c3e51 in ?? ()
from /usr/lib/x86_64-linux-gnu/libswscale.so.2
#1 0x00007ffff31a9e16 in sws_scale ()
from /usr/lib/x86_64-linux-gnu/libswscale.so.2
#2 0x00007ffff7bae851 in CvCapture_FFMPEG::retrieveFrame(int, unsigned char**, int*, int*, int*, int*) ()
from /usr/lib/libopencv_highgui.so.2.4
#3 0x00007ffff7bae8fb in cvRetrieveFrame_FFMPEG ()
from /usr/lib/libopencv_highgui.so.2.4
#4 0x00007ffff7badc49 in CvCapture_FFMPEG_proxy::retrieveFrame(int) () from /usr/lib/libopencv_highgui.so.2.4
#5 0x00007ffff7bac607 in cv::VideoCapture::retrieve(cv::Mat&, int) () from /usr/lib/libopencv_highgui.so.2.4
#6 0x00007ffff7bac407 in cv::VideoCapture::read(cv::Mat&) ()
from /usr/lib/libopencv_highgui.so.2.4
#7 0x00000000004017a1 in main ()
(gdb) q
A debugging session is active.
Inferior 1 [process 30800] will be killed.
Quit anyway? (y or n) y
</pre>
```
## History
##### Alexander Shishkov on 2013-05-17 14:02
```
Thank you for reporting the issue. Could you add more details about your OS and FFMPEG versions? Does this code work with other videos?
- Assignee set to Brian TheLion
- Category set to highgui-video
```
##### Brian TheLion on 2013-05-17 16:26
```
Yes, example code works fine for other videos. Here's how I discovered the issue:
1) The example code was initially implemented on Ubuntu 12.10 using the distro's libopencv-dev (OpenCV v2.3.1-7) package. It worked fine.
2) Last week I upgraded from Ubuntu 12.10 to 13.04.
3) I rebuilt the example code against the distro's libopencv-dev package, now because of the upgrade at OpenCV v2.4.2.
4) The bug appeared.
5) I uninstalled libopencv-dev and built OpenCV v2.4.5 from the source tarball.
6) I rebuilt the example code and still found the bug.
I had someone else here in the office verify this scenario as well. Thus, the bug appears to have been introduced somewhere between v2.3.1-7 and v2.4.2.
```
##### Brian TheLion on 2013-05-17 18:29
```
FFMPEG-related functionality on Ubuntu 13.04 is provided by libavutil. Here are the relevant dependencies from apt:
<pre>
br@breach:~$ aptitude show libopencv-highgui2.4
Package: libopencv-highgui2.4
State: installed
Automatically installed: yes
Version: 2.4.2+dfsg-0exp2ubuntu1
Priority: optional
Section: universe/libs
Maintainer: Ubuntu Developers <[email protected]>
Architecture: amd64
Uncompressed Size: 313 k
Depends: libavcodec53 (>= 6:0.8.3-1~) | libavcodec-extra-53
(>= 6:0.8.4), libavformat53 (>= 6:0.8.3-1~) |
libavformat-extra-53 (>= 6:0.8.4), libavutil51 (>=
6:0.8.3-1~) | libavutil-extra-51 (>= 6:0.8.4), libc6
(>= 2.15), libdc1394-22, libgcc1 (>= 1:4.1.1),
libglib2.0-0 (>= 2.31.8), libgtk2.0-0 (>= 2.8.0),
libilmbase6 (>= 1.0.1), libjasper1, libjpeg8 (>= 8c),
libopencv-core2.4, libopencv-imgproc2.4 (=
2.4.2+dfsg-0exp2ubuntu1), libopenexr6 (>= 1.6.1),
libpng12-0 (>= 1.2.13-4), libstdc++6 (>= 4.1.1),
libswscale2 (>= 6:0.8.3-1~) | libswscale-extra-2 (>=
6:0.8.4), libtiff5 (> 4.0.0-1~), libv4l-0 (>= 0.5.0)
</pre>
And the versions of the libav dependencies:
libavcodec53 6:0.8.6-1ubuntu2
libavcodec-extra-53 6:0.8.6ubuntu2
libavformat53 6:0.8.6-1ubuntu2
libavformat-extra-53 6:0.8.6ubuntu2
libavutil51 6:0.8.6-1ubuntu2
libavutil-extra-51 6:0.8.6ubuntu2
libswscale2 6:0.8.6-1ubuntu2
libswscale-extra-2 6:0.8.6ubuntu2
```
##### Brian TheLion on 2013-05-22 14:53
```
Can someone upstream please confirm this bug? I am now blocking on it. I have provided the code, the configuration, and an example video for which it fails. Cheers!
```
##### Jaime Alemany on 2013-08-17 00:32
```
Hi
I recently upgrade to ubuntu 13.04 and had the same problem. I resolved this setting the frame position to 0 before reading first frame with cvSetCaptureProperty().
After that the video plays again but now i found a new problem: some of the frames show corrupted data... (im trying to find why)
```
| bug,auto-transferred,priority: normal,category: videoio,affected: 2.4 | low | Critical |
97,425,048 | opencv | Mask threshold inconsistent over pyramids in orb detector | Transferred from http://code.opencv.org/issues/3037
```
|| Timo Breuer on 2013-05-19 12:45
|| Priority: Normal
|| Affected: 2.4.5 (latest release)
|| Category: features2d
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## Mask threshold inconsistent over pyramids in orb detector
```
The user provided mask for the feature detector is scaled the same way as the image, but additional thresholding is done.
@threshold(maskPyramid[level], maskPyramid[level], 254, 0, THRESH_TOZERO);@
The value of 254 sets all entries below to zero, which is a problem if the mask has not been created with entries of 0 and 255. Using a mask with entries of 0 and 1 yields keypoints on level 0 only, so the user is left with an error that is hard to track down.
Possible fixes: Set the threshold to 1 or set all non-zero entries of the original mask to 255.
```
## History
| bug,auto-transferred,priority: normal,affected: 2.4,category: features2d | low | Critical |
97,425,103 | opencv | GenericIndex lsh parameters wrong | Transferred from http://code.opencv.org/issues/3038
```
|| caron lee on 2013-05-20 02:14
|| Priority: High
|| Affected: 2.4.5 (latest release)
|| Category: flann
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## GenericIndex lsh parameters wrong
```
Code:
cvflann::LshIndexParams params;
pIndex = new GenericIndex<D_L2_float>(feature, params);
I tried the default LshIndexParams constructor or with parameters: LshIndexParams params(xx,xx,xx), but all those lead to a exception in :
modules/flann/include/opencv2/flann/params.h:63
the line is :
return it->second.cast<T>();
When I used gdb to get more information, I found that in the function get_param [modules/flann/include/opencv2/flann/params.h:59]
the params variable was right:
{
...
["key_size"] = {policy = 0xbdec68, object = 0x14},
["multi_probe_level"] = {policy = 0xbdec68, object = 0x2},
["table_number"] = {policy = 0xbdec68, object = 0xc}
}
but when I print the 'it' variable, I got:
p it
$16 = {first = "table_number", second = {policy = , object = }}
the second attribute was missing, and the it->second raised exception.
```
## History
##### Andrey Pavlenko on 2013-05-22 07:07
```
- Start date deleted (2013-05-20)
- Priority changed from Blocker to High
- Assignee set to Marius Muja
- Category set to flann
```
| bug,auto-transferred,affected: 2.4,category: flann | low | Critical |
97,425,182 | opencv | Connected Components Labeling in CUDA is buggy | Transferred from http://code.opencv.org/issues/3042
```
|| Andrey Pavlenko on 2013-05-24 15:53
|| Priority: Low
|| Affected: branch '2.4' (2.4-dev)
|| Category: gpu (cuda)
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Windows
```
## Connected Components Labeling in CUDA is buggy
```
The attached code results in incorrect image.
Also there is no test for it, if I'm not mistaken...
```
## History
##### Vadim Pisarevsky on 2015-05-21 22:19
```
ok, connected component is not the most important function which has to be run on GPU; considering the amount of available support and new Marina's projects, let's lower the priority
- Affected version changed from branch 'master' (2.4.9) to branch
'2.4' (2.4-dev)
- Operating System set to Windows
- HW Platform set to x64
- Priority changed from High to Low
```
| bug,auto-transferred,priority: low,affected: 2.4,category: gpu/cuda (contrib) | low | Critical |
97,425,378 | opencv | A "Preflight" Checklist | Transferred from http://code.opencv.org/issues/3054
```
|| Howard Pautz on 2013-05-29 23:55
|| Priority: Normal
|| Affected: None
|| Category: documentation
|| Tracker: Feature
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## A "Preflight" Checklist
```
Hope it's ok to suggest this here (thought it more appropriate than the forum)...
I'm an opencv n00b, but have years of C/C++ ...
Tried this week to build ocv for the first time. The goal was to run at least the 'show Lena' code. I've still not got it working ! ... tried on Fedora, Ubuntu, and Mac. All clean, up-to-date platforms. In addition to a simple GCC run, I also tried using the CDT/Eclipse approach on all three platforms. So that was 3 x 2 = 6 versions x 2 approaches = 12 failed attempts. The failures were at all stages for various reasons. I see the main problem as vague or incomplete documentation.
The docs for Mac/Linux need to explain where all the files go at the various stages of the build - I found little to no documentation on this crucial topic at all, but did come across many posts all over the net where people were looking for such locations. It would have saved me hours of headaches to have this at hand, as in a big, easy-to-see link from the home or doc home page:
Call it a preflight check list for getting Lena to show her lovely face. Some of this is 'sort of' explained, but not good enough if something fails. ( Got worn out using 'locate {filename}' )
0. put external libs we need {here} // e.g. libavcodec-dev, etc.
1. If you put git or tz file {some_dir}
2. Then unpack to {there}
3. You will find cmake files {somewhere_here_or_there}
4. After cmake, you will find logs {golly_only_knows}
5. The libs generated by cmake, such as libavcodec-dev, will go {here_lib}
6. Try running some sample code (... And *please* make sure it was built from scratch using this very check list ! )
7. If you get 'references not found' (unknown / undeclared) errors, then look back to {where_they_are} to make sure .h / .hpp are there.
8. If you get linker errors, or unknown references after an Eclipse build, look at {where_ever} to make sure .so files are there.
Happy ending with Lena :))
thanks
- Howard in Florida
```
## History
##### Alexander Smorkalov on 2013-05-30 06:01
```
- Description changed from Hope it's ok to suggest this here (thought
it more appropriate than the forum... to Hope it's ok to suggest
this here (thought it more appropriate than the forum... More
- Category set to documentation
```
##### Daniil Osokin on 2013-06-04 13:52
```
Hi, Howard!
Please, send a pull-request (http://code.opencv.org/projects/opencv/wiki/How_to_contribute) with your fixes: fix existing docs or create new with your instructions. It really speedups this process.
- Target version set to 2.4.6
- Assignee set to Howard Pautz
```
##### Kirill Kornyakov on 2013-06-14 09:45
```
Anybody who is ready to work on a new tutorial (or update the existing one), please follow http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute.
- Target version deleted (2.4.6)
```
| auto-transferred,priority: normal,feature,category: documentation | low | Critical |
97,425,441 | opencv | Access to OpenNI DeviceInfos for Multi-Sensor-Applications | Transferred from http://code.opencv.org/issues/3070
```
|| Hagen Borstell on 2013-06-04 13:36
|| Priority: Normal
|| Affected: None
|| Category: highgui-camera
|| Tracker: Feature
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## Access to OpenNI DeviceInfos for Multi-Sensor-Applications
```
Hello,
OpenNI provides the reading of device information (DeviceInfos: names, uri, pid, vendor, etc.). Thus, a sensor can be uniquely identified in a multi-sensor application.
This is important to clearly assign specific information (calibration parameters) to a specific sensor in multisensor applications.
Is it possible to make this information accessible over the OpenCV interface (cap_openni.cpp) or open a specific sensor (pid=....)?
p.s. In OpenCV you are limited to open "any" sensor --> VideoCapture capture( CV_CAP_OPENNI ); The order is arbitrary/random, right?
Thank you very much for your response.
```
## History
##### Daniil Osokin on 2013-06-04 13:47
```
Hi, Hagen!
Thank you for attention! Could you please contribute this functionality (http://code.opencv.org/projects/opencv/wiki/How_to_contribute)? OpenCV's developers will review it and add to OpenCV.
- Target version set to 2.4.6
- Assignee changed from Vadim Pisarevsky to Hagen Borstell
```
##### Kirill Kornyakov on 2013-06-14 09:50
```
- Target version deleted (2.4.6)
```
| auto-transferred,priority: normal,feature,category: videoio(camera) | low | Minor |
97,425,533 | opencv | VideoCapture does not support the Minoru 3D camera (and possibly other multi USB camera configurations) | Transferred from http://code.opencv.org/issues/3082
```
|| Matias Nitsche on 2013-06-11 18:57
|| Priority: Normal
|| Affected: 2.4.5 (latest release)
|| Category: highgui-video
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## VideoCapture does not support the Minoru 3D camera (and possibly other multi USB camera configurations)
```
Currently, it is not possible to instantiate two VideoCapture's using both USB devices corresponding to the Minoru 3D camera. According to what I've read, this is probably due to the fact that one camera takes up all available bandwidth. There's an existing implementation, libv4l2cam, which supports accessing the Minoru correctly.
I don't have the expertise to understand why this library works, but maybe someone with the knowledge may be able to change OpenCV VideoCapture code to support this kind of devices.
```
## History
| wontfix,auto-transferred,category: videoio,affected: 2.4 | low | Critical |
97,425,643 | opencv | The regression test Features2d_SURF_using_mask fails | Transferred from http://code.opencv.org/issues/3087
```
|| Ivan Korolev on 2013-06-14 13:21
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.4
|| Category: features2d
|| Tracker: Bug
|| Difficulty: None
|| PR:
|| Platform: None / None
```
## The regression test Features2d_SURF_using_mask fails
```
The @featureDetector_->detect(image, keyPoints, mask)@ method returns key points that are outside the mask.
```
## History
| bug,auto-transferred,priority: normal,affected: 2.4,category: features2d | low | Critical |
97,425,746 | opencv | Package a maven version of Java bindings | Transferred from http://code.opencv.org/issues/3097
```
|| Abhishek Kumar on 2013-06-17 16:00
|| Priority: Normal
|| Affected: None
|| Category: java bindings
|| Tracker: Feature
|| Difficulty:
|| PR:
|| Platform: None / None
```
## Package a maven version of Java bindings
```
We need to get OpenCV bundled with native code as a maven dependency.
Ant build systems have long been moving into oblivion. Though not getting deprecated anytime soon, all non-trivial have been moving towards Maven.
As such, availability of OpenCV as a maven dependency would satisfy many users while adding many new hobbyists to OpenCV user list.
```
## History
##### Daniel Baggio on 2014-01-09 21:00
```
Hi,
I have a sample pom.xml file that shows how to create a Java OpenCV project with Maven at:
https://github.com/JavaOpenCVBook/code/tree/master/chapter1/maven-sample
I would be glad to help in providing an official way to do it. Basic need is to generate a native JNI dynamic link library for each of the main platforms (windows, linux, osx, 32 and 64 bit). If that is created, a maven project is straightfoward.
Kind regards,
Daniel
```
##### Maurice Betzel on 2014-01-10 20:04
```
I have succesfully used the https://code.google.com/p/cmake-maven-project CMake-Maven plugin for cross platform deployments. Even have it running as a lib bundle under OSGi.
```
##### Thomas Weißschuh on 2014-03-12 21:21
```
+1
```
##### Daniil Osokin on 2014-03-17 07:41
```
Hi, guys! Please contribute this feature to OpenCV (code.opencv.org/projects/opencv/wiki/How_to_contribute). It will help a lot (and speed up the process in many times: OpenCV Jedis can efficiently communicate with you)!
```
##### Daniel Baggio on 2014-05-27 01:36
```
Actually, it doesn't have to do with the code, but it should address people who create the build itself.
With buildbot, I believe we should be able to get it working easily, but I don't know how to get in contact with people who has got to do with it.
```
##### Sergey Lisovoy on 2014-08-06 10:46
```
I create project on GitHub that build maven artifact that contains https://github.com/lis0x90/opencv-maven-bundle:
* opencv native (currently for linux64, win64 and win32)
* opencv java binding
* loader (use hawtjni to unpack and load native library in runtime)
Feel free to clone, build and test!
```
##### Daniel Baggio on 2014-10-21 03:38
```
Seriously, this should really be seen as a top priority task for Java developers.
Others also required it here:
http://answers.opencv.org/question/30945/why-isnt-the-opencv-java-libraries-available-via/
Any updates on this feature?
```
##### Steven Puttemans on 2014-10-21 07:53
```
Hi Daniel,
As long as noone will actually take up the task and do it, then I am afraid it could take quite a long time since OpenCV developers are doing this in their spare time.
Are you up to supplying a PR for this yourself? Or contact the right people for it?
Steven
```
##### Daniel Baggio on 2014-11-12 23:04
```
Sorry, Steven, but what do you mean by PR?
Kind regards,
Daniel
```
##### Steven Puttemans on 2014-11-13 13:32
```
With a pull request I mean an adaptation done by you to the code base through the github interface.
More information can be found here: http://code.opencv.org/projects/opencv/wiki/How_to_contribute
```
##### Daniel Baggio on 2014-11-19 01:53
```
Thanks Steven.
Unfortunately, this can't be done with a pull request since this is part of the build process. If only could I get in touch with people who make OpenCV releases, this could be addressed more easily. It is just a matter of uploading jar files to maven repositories, that would make a huge difference for Java developers, since it could take away the part of generating binaries.
```
##### Stojan Pesov on 2015-03-04 08:42
```
I agree with Daniel, we are eager to see the jar put in maven from you guys officially.
That will be great advantage for all Java developers.
```
##### Daniel Baggio on 2015-06-25 05:56
```
Please, has anyone seen this feature?
```
| auto-transferred,feature,priority: low,category: java bindings | low | Critical |
97,425,893 | opencv | Stitcher generate result file 158 bytes long without error | Transferred from http://code.opencv.org/issues/3122
```
|| Nikolay Rogoshchenkov on 2013-06-28 21:22
|| Priority: Normal
|| Affected: branch 'master' (2.4.9)
|| Category: stitching
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Windows
```
## Stitcher generate result file 158 bytes long without error
```
*Subj!*
Command line to reproduce:
<pre>
openCV_stitch_three.exe b1.bmp b2.bmp b3.bmp
</pre>
Example code:
<pre>
int main(int argc, char* argv[])
{
int retval = parseCmdArgs(argc, argv);
if (retval) return -1;
Mat pano;
printf("Channels image1: %d\n", imgs[0].type());
printf("Channels image2: %d\n", imgs[1].type());
int width = 170;
cv::Rect rect1 = cvRect( imgs[0].cols-width, 0, width, imgs[0].rows); //second half of the first image
cv::vector<cv::Rect> roi1;
roi1.push_back(rect1);
cv::Rect rect12 = cvRect(0, 0,width, imgs[1].rows); //first half of the second image
cv::Rect rect23 = cvRect(imgs[1].cols -width, 0, width, imgs[1].rows); //second half of the first image
cv::vector<cv::Rect> roi2;
roi2.push_back(rect12);
roi2.push_back(rect23);
cv::Rect rect3 = cvRect(0, 0, width, imgs[2].rows); //first half of the third image
cv::vector<cv::Rect> roi3;
roi3.push_back(rect3);
cv::vector<cv::vector<cv::Rect>> rois;
rois.resize(3);
rois[0] = roi1;
rois[1] = roi2;
rois[2] = roi3;
//cv::imshow("debug_img1", imgs[0]);
Mat tmp1, tmp2, tmp3;
cvtColor(imgs[0], tmp1, CV_BGR2GRAY);
cvtColor(imgs[1], tmp2, CV_BGR2GRAY);
cvtColor(imgs[2], tmp3, CV_BGR2GRAY);
/*lensCorrectorBarrel(tmp1.data, tmp1.cols, tmp1.rows);
lensCorrectorBarrel(tmp2.data, tmp2.cols, tmp2.rows);
lensCorrectorBarrel(tmp3.data, tmp3.cols, tmp3.rows);
imwrite("corr1.tif", tmp1);
imwrite("corr2.tif", tmp2);
imwrite("corr3.tif", tmp3);*/
/*namedWindow( "debug_img_corr1", CV_WINDOW_AUTOSIZE );
namedWindow( "debug_img_corr2", CV_WINDOW_AUTOSIZE );
cv::imshow("debug_img_corr1", tmp1);
cv::imshow("debug_img_corr2", tmp2);*/
cvtColor(tmp1, imgs[0], CV_GRAY2BGR);
cvtColor(tmp2, imgs[1], CV_GRAY2BGR);
cvtColor(tmp3, imgs[2], CV_GRAY2BGR);
//char cc = cvWaitKey(0);
/*cv::rectangle( imgs[0], rect1, CV_RGB(255,0,0), 1);
cv::rectangle( imgs[1], rect12, CV_RGB(255,0,0), 1);
cv::rectangle( imgs[1], rect23, CV_RGB(255,0,0), 1);
cv::rectangle( imgs[2], rect3, CV_RGB(255,0,0), 1);
cv::imshow("debug_img1", imgs[0]);
cv::imshow("debug_img2", imgs[1]);
cv::imshow("debug_img3", imgs[2]);
char c = cvWaitKey(0);*/
Stitcher stitcher = Stitcher::createDefault(false);
stitcher.setPanoConfidenceThresh(0.07);
stitcher.setWaveCorrection(false);
//detail::BestOf2NearestMatcher *matcher = new detail::BestOf2NearestMatcher(false, 0.001/*=match_conf*/);
//detail::SurfFeaturesFinder *featureFinder = new detail::SurfFeaturesFinder(100);
//stitcher.setFeaturesMatcher(matcher);
//stitcher.setFeaturesFinder(featureFinder);
Stitcher::Status status = stitcher.stitch(imgs, rois, pano);
//Stitcher::Status status = stitcher.composePanorama(imgs, pano);
if (status != Stitcher::OK)
{
cout << "Can't stitch images, error code = " << status << endl;
return -1;
}
imwrite(result_name, pano);
char ac = cvWaitKey(0);
if(ac == 'ESC')
return 0;
}
</pre>
Thank you.
```
## History
##### Nikolay Rogoshchenkov on 2013-06-28 21:27
```
How can I attach additional files? With good images, for example.
```
##### Alexey Spizhevoy on 2013-07-01 06:27
```
There is the 'Files' section available when one is composing a new message.
```
##### Nikolay Rogoshchenkov on 2013-07-01 12:12
```
Since archieve with good images > 5MB, you can download it here: http://nikkadim.com/good_images.tar.gz ~6.9MB
Thank you.
```
##### Alexey Spizhevoy on 2013-07-02 08:32
```
The code works fine when the images are passed in the correct order. Otherwise the application generates trash (as expected), but sometimes even without error message -- which isn't OK. I will add an error notification.
```
##### Nikolay Rogoshchenkov on 2013-07-04 22:25
```
Alexey Spizhevoy wrote:
> The code works fine when the images are passed in the correct order. Otherwise the application generates trash (as expected), but sometimes even without error message -- which isn't OK. I will add an error notification.
So, you wanna say, that's order in my example is wrong?
```
##### Alexey Spizhevoy on 2013-07-05 07:52
```
If I pass "a1.bmp a2.bmp a3.bmp" (you haven't attached any b*.bmp) then your code works, in some other cases it fails.
- Affected version changed from 2.4.5 (latest release) to branch
'master' (2.4.9)
```
##### Nikolay Rogoshchenkov on 2013-07-05 19:19
```
This problem.tar.gz archive contains:
<pre>
- openCV_stitch_three.exe
- b1.bmp
- b2.bmp
- b3.bmp
- result.tif
</pre>
Command line to reproduce:
<pre>
openCV_stitch_three.exe b1.bmp b2.bmp b3.bmp
</pre>
In this case we have result.tif with size 158 bytes (header).
Thank you.
- File problem.tar.gz added
```
##### Alexey Spizhevoy on 2013-07-09 08:21
```
I've checked: "openCV_stitch_three.exe b1.bmp b2.bmp b3.bmp" failed, but "openCV_stitch_three.exe b1.bmp b3.bmp b2.bmp" generated good panorama.
```
##### Nikolay Rogoshchenkov on 2013-07-11 02:37
```
Alexey Spizhevoy wrote:
> I've checked: "openCV_stitch_three.exe b1.bmp b2.bmp b3.bmp" failed, but "openCV_stitch_three.exe b1.bmp b3.bmp b2.bmp" generated good panorama.
Failed - you mean you've got error? Or 158 bytes result file?
But problem is the right sequence is b1.bmp b2.bmp b3.bmp, this images from 3 camera in one row.
```
##### Nikita Manovich on 2013-07-11 07:56
```
- Status changed from New to Open
```
##### Alexey Spizhevoy on 2013-07-11 07:59
```
> Failed - you mean you've got error? Or 158 bytes result file?
158 bytes result.
> But problem is the right sequence is b1.bmp b2.bmp b3.bmp, this images from 3 camera in one row.
Maybe, but it turns out that b1,b3,b2 are matched better, than b1,b2,b3. The images' texture is very specific.
```
##### Nikolay Rogoshchenkov on 2013-07-11 14:41
```
Alexey Spizhevoy wrote:
> > Failed - you mean you've got error? Or 158 bytes result file?
> 158 bytes result.
>
> > But problem is the right sequence is b1.bmp b2.bmp b3.bmp, this images from 3 camera in one row.
> Maybe, but it turns out that b1,b3,b2 are matched better, than b1,b2,b3. The images' texture is very specific.
Ok, is it normal to receive 158 byte file without error from stitching module? Probably not.
```
##### Alexey Spizhevoy on 2013-07-12 08:57
```
I've already said before that silent stitching failure is not OK. The "pull request":https://github.com/Itseez/opencv/pull/1079 is waiting to be merged.
```
##### Nikolay Rogoshchenkov on 2013-07-12 11:39
```
Thank you! Sorry, I did not notice that.
This update will be available only in 2.4.9, right?
```
##### Nikolay Rogoshchenkov on 2013-07-12 13:47
```
Alexey, i think the problem with 158 bytes size file is more deeper.
I've compiled OpenCV246 with CUDA and TBB support, and wrote a simple program (see attachment) for stitching 6 images.
But each time when I start this program I've got different results:
!http://nikkadim.com/stitch_six.png!
Each time when stitching tooks >10 secs I had result image size 158 bytes, than I restart program with the same input image data I had a good results (with calculation time ~5 sec).
- File main.cpp added
```
##### Nikolay Rogoshchenkov on 2013-07-12 13:48
```
input files available here: http://nikkadim.com/input_imges.zip ~5.5Mb
```
##### Alexey Spizhevoy on 2013-07-12 14:27
```
The fix will be available in the next major release. What version the master branch will become isn't determined yet. But the issue with error message isn't critical anyway, however if you need the fix you can build OpenCV from the master branch when the pull request is merged.
As for the new info, do you have the same problem without CUDA support?
```
##### Nikolay Rogoshchenkov on 2013-07-13 21:19
```
Alexey Spizhevoy wrote:
> As for the new info, do you have the same problem without CUDA support?
Thank you, Alexey!
I did not try, but anyway, I did not use CUDA instructions in program, you can see it. Should I anyway try to compile OpenCV without CUDA?
```
##### Alexey Spizhevoy on 2013-07-15 12:09
```
I couldn't reproduce the problem. But I guess the behavior you observed was due to the FLANN matcher (used in BestOf2NearestMatcher) and it's non-deterministic actually.
```
##### Nikolay Rogoshchenkov on 2013-07-15 22:01
```
Alexey Spizhevoy wrote:
> I couldn't reproduce the problem. But I guess the behavior you observed was due to the FLANN matcher (used in BestOf2NearestMatcher) and it's non-deterministic actually.
If needed I can share the dll's, exe and input files to reproduce the issue.
Also I have noticed, that this issue also reproduced on my desktop PC (Intel Core i7 950/24GB DDR3/Win7 x64). I can reproduce this with 32-bit and 64-bit version of program.
How can I also help to catch the root of case? May try to play with BestOf2NearestMatcher params?
Thank you, Alexey.
```
##### Nikolay Rogoshchenkov on 2013-10-30 20:00
```
any updates?
```
##### Nikolay Rogoshchenkov on 2015-04-07 16:19
```
Same with 2.4.11 for both x86 and x64 arch's.
Resulting pano has size = (1,1), but calculation time about right
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: stitching | low | Critical |
97,426,143 | opencv | FileStorage can't read VOC annotations | Transferred from http://code.opencv.org/issues/3154
```
|| Ming-Ming Cheng on 2013-07-12 13:23
|| Priority: Normal
|| Affected: None
|| Category: core
|| Tracker: Feature
|| Difficulty:
|| PR:
|| Platform: None / None
```
## FileStorage can't read VOC annotations
```
VOC and ImageNet use the same xml annotation format. Although they are most famous computer vision dataset. The annotation format is not supported in OpenCV. Some reasons for crash includes:
1. The VOC annotation format don't start with '<?xml version="1.0"?> <opencv_storage>'.
2. Some values, e.g. <owner><flickrid>, might contains characters not supported.
3. Values like <filename>000005.jpg</filename> needs to be changed to <filename>"000005.jpg"</filename> in order to avoid crash in reading.
Above problems can be avoid by some simple pre-processing. The other bug, which is reported by OpenCV as "Duplicated key", can't be easily addressed. However, each VOC image has several <object> key.
```
## History
##### Nikita Manovich on 2013-07-12 14:30
```
Hi,
Thank you for the report. I don't think that it is a bug. You have suggested some new functionality. Could you please help us and submit a pull request? Unfortunately we have limited resources. We will be glad to see you as a contributer. Please read http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute for more information.
- Difficulty deleted (Medium)
- Assignee set to Maria Dimashova
- Status changed from New to Open
- Category set to core
- Tracker changed from Bug to Feature
```
##### Nikita Manovich on 2013-07-12 14:30
```
- Assignee deleted (Maria Dimashova)
```
| auto-transferred,priority: normal,feature,category: core | low | Critical |
97,426,184 | opencv | Support disabling LZW compression for writing TIFF images | Transferred from http://code.opencv.org/issues/3158
```
|| Nikolay Rogoshchenkov on 2013-07-16 12:28
|| Priority: Normal
|| Affected: None
|| Category: highgui-images
|| Tracker: Feature
|| Difficulty:
|| PR:
|| Platform: None / None
```
## Support disabling LZW compression for writing TIFF images
```
Will be very useful to have ability to disable/enable LZW compression for writing TIFF-files. Now TIFF-files writing with default LZW compression.
```
## History
##### Nikita Manovich on 2013-07-16 18:24
```
Hi,
Thank you for the feature request. If you can help us and implement the functionality it will be cool. You can find all necessary information on the following page: http://code.opencv.org/projects/opencv/wiki/How_to_contribute.
- Status changed from New to Open
- Category set to highgui-images
```
| auto-transferred,priority: normal,feature,category: imgcodecs | low | Minor |
97,426,222 | opencv | BORDER_WRAP not available in FilterEngine | Transferred from http://code.opencv.org/issues/3168
```
|| Tama McGlinn on 2013-07-20 23:44
|| Priority: Low
|| Affected: None
|| Category: imgproc, video
|| Tracker: Feature
|| Difficulty: Easy
|| PR:
|| Platform: None / None
```
## BORDER_WRAP not available in FilterEngine
```
modules/imgproc/src/filter.cpp:160 - CV_Assert( columnBorderType != BORDER_WRAP );
This line prevents me from making a linear filter with BORDER_WRAP type. I have no idea why this is, nor how to fix it.
```
## History
##### Steven Puttemans on 2013-07-21 12:25
```
This is due to no code in the filter algorithm being provided for solving problems with the cv_borderwrap type of borderstyle. To fix it what you can do is create a pull request with the assert gone, then see how it influences the problem and where it goes wrong. But I am guessing some algorithms beneath it will generate errors which need to be fixed then.
```
##### Ivan Korolev on 2013-07-22 10:35
```
Hello Tama McGlinn,
Thank you for submitting this ticket. Unfortunately, FilterEngine doesn't support BORDER_WRAP type of border. If you could add this feature on your side, a pull request in our GitHub repo (http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute for more details) would be highly appreciated! If you decide to continue with introducing this feature then please explain what is purpose of this feature?
- Priority changed from Normal to Low
- Status changed from New to Open
- Tracker changed from Bug to Feature
```
##### Alexander Smorkalov on 2013-11-27 10:53
```
- Target version changed from 2.4.7 to Next Hackathon
- Assignee deleted (Vadim Pisarevsky)
```
##### isscandar isscandar on 2014-09-30 18:10
```
Ivan Korolev wrote:
> If you decide to continue with introducing this feature then please explain what is purpose of this feature?
Hello Ivan,
in my applications I often work with a region of interest with the shape of an annulus, for some particular tasks I rectify the annulus transforming it in a rectangular stripe; a circular path (ideally one-pixel wide) inside the annulus is so transformed into a row in the rectangular stripe, for these rows I think that the @BORDER_WRAP@ interpolation would be the best interpolation scheme. I faced the same need of Tama McGlinn when I tried to do @cv::cornerHarris@ with @borderType@ equal to @cv::BORDER_WRAP@ (OpenCV 2.4.4).
Thank you.
```
| auto-transferred,feature,category: imgproc,category: video,priority: low | low | Critical |
97,426,257 | opencv | BFMatcher | Transferred from http://code.opencv.org/issues/3172
```
|| Robert Aschenbrenner on 2013-07-24 17:49
|| Priority: Normal
|| Affected: 2.4.9 (latest release)
|| Category: features2d
|| Tracker: Bug
|| Difficulty:
|| PR: https://github.com/Itseez/opencv/pull/3025
|| Platform: x64 / Linux
```
## BFMatcher
```
Using BFMatcher with "NORM_HAMMING" on binary descriptors (created by FREAK) gives unexpected results in certain conditions:
<pre>
std::vector<Mat> descriptors; //FREAK descriptors of e.g. 1000 images
... //compute actual descriptors
BFMatcher bfMatcher(NORM_HAMMING, false);
bfMatcher.add(descriptors);
int nn = 10; //Number of nearest neighbours
bfMatcher.knnMatch(query_descriptors, matches, nn);
</pre>
The above code returns invalid results, if *nn* happens to be smaller than any image's descriptor set.
For example one single image _test.jpg_ might have only *9* (<nn) according FREAK descriptors.
The BFMatcher will calculate all distances using @batchDistance@ (https://github.com/Itseez/opencv/blob/master/modules/features2d/src/matchers.cpp#L366).
Eventually it passes the current best results (@nidx@) to @cv::batchDistance@ in https://github.com/Itseez/opencv/blob/master/modules/core/src/stat.cpp#L1792.
The problem arises in the following section:
<pre>
K = std::min(K, src2.rows);
_dist.create(src1.rows, (K > 0 ? K : src2.rows), dtype);
Mat dist = _dist.getMat(), nidx;
if( _nidx.needed() )
{
_nidx.create(dist.size(), CV_32S);
nidx = _nidx.getMat();
}
</pre>
@K@ is reduced, when _test.jpg_ (src2) is inspected, since it only has *9* rows.
@_nidx.create(dist.size(), CV_32S);@ reduces the dimension of @nidx@ to 9 columns, invalidating older results.
Later images may increase the column size again to 10.
The assumption, that the number of descriptors of every image is always > nn, seems unenforced.
```
## History
##### Ivan Korolev on 2013-07-25 06:43
```
Hi Robert,
thanks for the bug report. Do you think you can fix the problem? If you could investigate this issue or send a fixing pull request (http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute for more details) your help would be highly appreciated!
```
##### Ivan Korolev on 2013-07-25 06:43
```
- Status changed from New to Open
```
##### Robert Aschenbrenner on 2013-08-03 13:18
```
Hi, I am afraid that I do not have sufficient time at the moment. In my code I'm circumventing the above mentioned case by extra checks.
I suspect the code in stat.cpp is not only used by BFMatcher.
Maybe the original author could have a look?
```
##### Philipp Hasper on 2014-08-04 12:28
```
Unfortunately it seems that I don't have the rights to assign a pull request to this issue. So I have to paste the URL manually: https://github.com/Itseez/opencv/pull/3025
```
##### Alexander Karsakov on 2014-08-14 11:47
```
- Affected version changed from 2.4.0 - 2.4.5 to 2.4.9 (latest
release)
- Pull request set to https://github.com/Itseez/opencv/pull/3025
- Assignee deleted (Maria Dimashova)
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: features2d | low | Critical |
97,426,308 | opencv | cv:HoGDescriptor support for 16bit gray-scale images | Transferred from http://code.opencv.org/issues/3175
```
|| Siegfried Hochdorfer on 2013-07-25 10:03
|| Priority: Normal
|| Affected: None
|| Category: objdetect
|| Tracker: Feature
|| Difficulty:
|| PR:
|| Platform: None / None
```
## cv:HoGDescriptor support for 16bit gray-scale images
```
Hi,
the current implementation of the cv::HoGDescriptor supports CV_8U and CV_8UC3 images. I am very interested in an implementation of the cv::HoGDescriptor for 16bit gray-scale (CV_16U) images. I want to try to implement it. But before I start I have two questions. First, is it in generally possible to implement support for 16U_C1 images? Second, how (and who) can I ask for help if i get in problems while trying to implement 16bit support?
```
## History
##### Ivan Korolev on 2013-07-26 06:14
```
Hi Siegfried,
thank you for your participation, this functionality will be useful!
> First, is it in generally possible to implement support for 16U_C1 images?
I think it is possible.
> Second, how (and who) can I ask for help if i get in problems while trying to implement 16bit support?
All your questions can be asked here or on GitHub (for more details http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute). The second variant is more preferable, because GitHub is the best place to discuss code.
- Assignee set to Marina Kolpakova
- Status changed from New to Open
```
| auto-transferred,priority: normal,feature,category: objdetect | low | Minor |
97,426,351 | opencv | Add EPNP and P3P to cvFindExtrinsicCameraParams2 | Transferred from http://code.opencv.org/issues/3178
```
|| Goncalo Lopes on 2013-07-27 11:45
|| Priority: Normal
|| Affected: None
|| Category: calibration, 3d
|| Tracker: Feature
|| Difficulty:
|| PR:
|| Platform: None / None
```
## Add EPNP and P3P to cvFindExtrinsicCameraParams2
```
It would be extremely helpful to have the EPNP method on the C API as well. It should be relatively easy to add a "flags" parameter at the end with a default value to keep backwards compatibility and use internal conversions to cv::Mat to use solvePnP.
I know the C API is not currently under active development, but for some applications it is still unfortunately the only effective way to interop with OpenCV...
```
## History
##### Steven Puttemans on 2013-07-28 15:15
```
I guess this is what you call a lost cause. The old OpenCV 1.x C API is depricated, support and development on it has been removed and most likely it will be removed completely in the openCV 3.0 interface. So I don't think more C code will be added to the base system.
- Assignee set to Alexander Shishkov
```
##### Victor Kocheganov on 2013-08-06 08:58
```
Hello Goncalo,
thank you for sharing your ideas and submitting this ticket!
Unfortunately our resources are very limited at the moment. So if you will have time to help the community and implement this feature by yourself that would be great and highly appreciated! (please read http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute for more details).
Thank you in advance,
Victor Kocheganov
- Target version set to 2.4.7
- Assignee changed from Alexander Shishkov to Vadim Pisarevsky
- Category set to calibration, 3d
```
##### Victor Kocheganov on 2013-08-08 05:41
```
- Target version changed from 2.4.7 to 3.0-alpha
- Status changed from New to Open
```
##### Maksim Shabunin on 2015-04-27 09:11
```
- Target version changed from 3.0-alpha to 3.0
```
| auto-transferred,priority: normal,feature,category: calib3d | low | Minor |
97,426,573 | opencv | white border while displaying a full image with python2.7 and opencv | Transferred from http://code.opencv.org/issues/3188
```
|| Stefano Spirolazzi on 2013-08-02 10:49
|| Priority: Normal
|| Affected: branch '2.4'
|| Category: highgui-images
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x86 / Any
```
## white border while displaying a full image with python2.7 and opencv
```
Hi!
I want to display a black image in a full screen mode. It looks a really sample task to do, but when i run the code below i get a little white strip on top and on the left. It doesn't matter if i change the size of the numpy array, i always get it.
If you run this code, do you get a fully black screen??
I have tried with different laptop and different version of opencv
thanks
@img = np.zeros((800, 1280)) #my resolution is 800, 1280
cv2.namedWindow("test", cv2.WND_PROP_FULLSCREEN)
cv2.setWindowProperty("test", cv2.WND_PROP_FULLSCREEN, cv2.cv.CV_WINDOW_FULLSCREEN)
cv2.imshow("test",img)
cv2.waitKey(0)@
```
## History
##### Victor Kocheganov on 2013-08-06 06:55
```
Hello Stefano,
thank you for submitting this ticket!
Could you please duplicate this question to http://answers.opencv.org/questions/, likely somebody's already faced with similar issue and could help. And it would be highly appreciated if you'll have time to investigate this issue and propose your own fix (if this is indeed a bug). Please see also http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute for details.
Thank you in advance,
Victor Kocheganov
- Target version set to 2.4.7
- Assignee set to Vadim Pisarevsky
- Status changed from New to Open
- Category set to highgui-images
```
##### Victor Kocheganov on 2013-08-09 10:37
```
- Assignee changed from Vadim Pisarevsky to Alexander Smorkalov
```
##### Alexander Smorkalov on 2013-08-29 13:59
```
Hello Stefano!
I tried to reproduce your issue on my Linux desktop. Your code works perfect for me. I use Ubuntu and GTK back-end for OpenCV. Please select proper platform in ticket properties if you use something other. I will try the same on Windows and close the issue if everything is ok.
My python and c++ code:
<pre>
import numpy as np
import cv2
img = np.zeros((1080, 1920)) #my resolution is 800, 1280
cv2.namedWindow("test", cv2.WND_PROP_FULLSCREEN)
cv2.setWindowProperty("test", cv2.WND_PROP_FULLSCREEN, cv2.cv.CV_WINDOW_FULLSCREEN)
cv2.imshow("test",img)
cv2.waitKey(0)
</pre>
<pre>
#include <opencv2/core/core.hpp>
#include <opencv2/highgui/highgui.hpp>
int main(int, char**)
{
cv::Mat frame = cv::Mat::zeros(1080, 1920, CV_8UC3);
cv::namedWindow("test", cv::WND_PROP_FULLSCREEN);
cv::setWindowProperty("test", cv::WND_PROP_FULLSCREEN, CV_WINDOW_FULLSCREEN);
cv::imshow("test", frame);
cv::waitKey(0);
return 0;
}
</pre>
```
##### Alexander Smorkalov on 2013-08-29 14:52
```
I reproduce the issue on Windows 8 with WIN32UI back-end.
```
##### Alexander Smorkalov on 2013-09-12 10:51
```
- Target version changed from 2.4.7 to Next Hackathon
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: imgcodecs | low | Critical |
97,426,768 | opencv | ffmpeg decoder crash for raw (uncompressed) avi file | Transferred from http://code.opencv.org/issues/3199
```
|| Guillaume Dumont on 2013-08-06 14:47
|| Priority: Normal
|| Affected: 2.4.6 (latest release)
|| Category: highgui-video
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x86 / Windows
```
## ffmpeg decoder crash for raw (uncompressed) avi file
```
In CvCapture_FFMPEG::grabFrame :
The call to avcodec_decode_video2 does not seem to allocate memory for the AVFrame that is passed to it when a raw uncompressed video is decoded. It seems like the call to av_free_packet frees the memory that the AVFrame::data[0] is pointing to.
then in CvCapture_FFMPEG::retrieveFrame :
the call to sws_scale crashes because ffmpeg tries to write to invalid memory.
Any idea how to fix this?
```
## History
##### Victor Kocheganov on 2013-08-08 04:55
```
Hello Guillaume Dumont,
thank you for reporting the issue!
Could you please provide some more information to reproduce the issue: example code, output logs.
Thanks.
- Status changed from New to Incomplete
- Category set to highgui-video
```
##### Guillaume Dumont on 2013-08-12 16:02
```
Hi,
Thanks for the response. Attached is the source code of two sample applications, one which uses ffmpeg 0.10.7 (which I think is the same one, test, that opencv uses) I compiled myself with MinGW and one, test2, that uses that one provided with opencv as a prebuilt binary. The crash occurs in both applications.
See comments main2.cpp line 19 and cap_ffmpeg_impl.hpp lines 684-685. If line 685 is commented no crash occurs when sws_scale is called in cap_ffmpeg_impl.hpp at line 728. To me it seems like the memory freed by av_free_packet should not be freed or there should be a copy when
avcodec_decode_video2 is called at line 657 in cap_ffmpeg_impl.hpp. Please note that no crash occurs if we use the VFW backend instead of ffmpeg.
The video I'm a using to test was too large to fit in the attachments and I'd rather not make it publicly available so let me know if I can transfer it to you by some other mean.
- File opencv_ffmpeg.7z added
```
##### Jaime Alemany on 2013-08-17 01:40
```
Hi
What about removing the last call to av_free_packet in grabFrame?
I have recently upgraded mi OS from Ubuntu 12 to 13.04. After that if i try to play some uncompressed avi videos sometimes have a crash, sometimes i can see corrupted images on the window. I have commented this call to the function and now seems to work fine (opencv 2.4.6)
I have not tested for memory leaks but seems that the memory used when playing very long videos is correctly managed.
```
##### Jaime Alemany on 2013-08-21 16:35
```
I can confirm that: after the call to av_decode_video2 the picture->data is the same pointer than packet.data. So, when calling av_free_packet this memory is freed and after thet it cannot be referenced as it is done in retrieveframe function. This is probably a bug (or a feature) in the ffmpeg decode function. Attached a ddd debuger screen capture.
Removing the last call to av_free_packet allows to call retrievefreme() without accessing the freed memory but, i found a problem when seeking a frame (cvSetCaptureProperty) in a YUYV video file. After seeking grabFrame returns 0 and frames cannot be read.
EDIT: opencv 2.4.6.1 compiled with ffmpeg support (last version of ffmpeg compiled from source 23-08-2013), works fine. Line 683 in cap_ffmpeg_impl.hpp commented (line 699 in the git version, call to av_free_packet) and no crashes, no access to freed memory (valgrind doesnt report) and no seek issues in video files with YUY2 codec.
- File picture.png added
```
| bug,auto-transferred,priority: normal,category: videoio,affected: 2.4 | low | Critical |
97,426,815 | opencv | VideoWriter output with VP8 not working on Fedora and Ubuntu | Transferred from http://code.opencv.org/issues/3203
```
|| Jan Sedlák on 2013-08-08 14:01
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.5
|| Category: highgui-video
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Linux
```
## VideoWriter output with VP8 not working on Fedora and Ubuntu
```
Hi, I have asked "this question":http://answers.opencv.org/question/18203/videowriter-output-using-vp8-not-working/ but no one answered it and I think that this is bug.
I want to use VideoWriter with webm output, I am using VP8 codec. It creates VideoWriter correctly and without an error. It creates output file and it puts information about resolution in it, but when I write image into it, it doesn't do anything and output file is always only 7 KiB. I have also tried it in C++ with the same result.
I am using Fedora 19, 64bit. I have added RPM Fusion repos and tried to install every gstreamer plugin, ffmpeg and everything that popped out when I searched "VP8". I also tried it in Ubuntu 13.04 with the same result. When I tried to use different codec (xvid), it worked.
You can try it with code I have pasted on answers.opencv.org.
```
## History
##### Alexander Smorkalov on 2013-08-08 18:16
```
- Target version set to Next Hackathon
- Assignee set to Alexander Smorkalov
- Category set to highgui-video
```
##### Victor Kocheganov on 2013-08-09 08:43
```
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,category: videoio,affected: 2.4 | low | Critical |
97,426,897 | opencv | gpu::StereoConstantSpaceBP crash | Transferred from http://code.opencv.org/issues/3211
```
|| Roman P on 2013-08-12 13:39
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.6
|| Category: gpu (cuda)
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Windows
```
## gpu::StereoConstantSpaceBP crash
```
When calling StereoConstantSpaceBP operator() on some images, OpenCV crashes with the error message:
OpenCV Error: Gpu API call (unknown error) in unknown function, file D:/Libraries/opencv_2_4_6/opencv/modules/gpu/src/cuda/stereocsbp.cu, line 683
System used: Windows 7 x64, NVIDIA GPU NVS 300, 8 GB RAM, CUDA 5.5, MS Visual Studio Professional 2012 (update 3)
Code reproducing this behavior and example images are attached.
```
## History
##### Steven Puttemans on 2013-08-12 14:38
```
I guess it is a stupid question, but you did rebuilt openCV to include GPU support right?
```
##### Roman P on 2013-08-13 12:24
```
Yes, OpenCV was build with the GPU support.
```
##### Kirill Kornyakov on 2013-08-14 12:00
```
OK, we'll try to check it on our side. Vlad, could you have a look please?
BTW, Roman, if you're able to debug the issue yourself, please work on a pull request. Vladislav is on vacations, and I'm not sure whether he is going to look at this issue soon...
- Target version set to 2.4.7
- Assignee set to Vladislav Vinogradov
- Category set to gpu (cuda)
```
##### Kirill Kornyakov on 2013-08-14 12:04
```
- Priority changed from High to Normal
```
##### Sergei Nosov on 2013-08-14 15:03
```
- Status changed from New to Open
```
##### Alexander Smorkalov on 2013-11-27 10:23
```
Vlad, could you have a look please?
- Affected version changed from 2.4.6 (latest release) to 2.4.0 -
2.4.6
```
##### Alexander Smorkalov on 2013-11-27 10:23
```
- Target version changed from 2.4.7 to 2.4.8
```
##### Alexander Smorkalov on 2013-12-30 10:38
```
- Target version changed from 2.4.8 to 2.4.9
```
##### Alexander Smorkalov on 2014-04-30 19:05
```
- Target version changed from 2.4.9 to 2.4.10
```
##### Andrea Tomassetti on 2015-03-25 11:39
```
Hi,
I know that this a very old issue but i have the same exactly problem. From time to time, with some images that have different sizes, when calling StereoConstantSpaceBP operator(), OpenCV crashes
<pre>
OpenCV Error: Gpu API call (an illegal memory access was encountered) in unknown function, file D:/toolkits/opencv-2.4.9/modules/gpu/src/cuda/stereocsbp.cu, line 683
error: D:/toolkits/opencv-2.4.9/modules/gpu/src/cuda/stereocsbp.cu:683: error: (-217) an illegal memory access was encountered
</pre>
System used: Windows 8.1 x64, NVIDIA GPU GeForce 740M, 8 GB RAM, CUDA 6.5, MS Visual Studio Professional 2013
The code that sometimes works and sometimes doesn't is the one tooked from gpu samples.
I can try debug the issue myself, if someone, kindly, show me the way.
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: gpu/cuda (contrib) | low | Critical |
97,426,968 | opencv | namedWindow approach in combination with waitKey | Transferred from http://code.opencv.org/issues/3218
```
|| Steven Puttemans on 2013-08-13 12:40
|| Priority: Normal
|| Affected: None
|| Category: highgui-gui
|| Tracker: Feature
|| Difficulty:
|| PR:
|| Platform: None / None
```
## namedWindow approach in combination with waitKey
```
For the moment, if you want your system to be able to wait for a keystroke, the waitKey() functionality can be used. However this requires you to have an active and open namedWindow. How hard would it be to remove this dependency? Many time I want to use the waitKey functionality, without opening a window, just by displaying results in the command line. I guess others will have experienced the same.
About the namedWindow, I think it is kind of weird that the namedWindow cannot be used like an object. Each function calling to this namedWindow, needs the exact name of the window again, which makes it kind of useless to predefine a namedWindow. Might it be possible to make an object from this, on which you can call a destroy for example?
```
## History
##### Sergei Nosov on 2013-08-14 13:27
```
Hi, Steven!
Your suggestions are sound, but I don't think the issues really have high exposure.
1) you can always use smth like getchar to wait for a keystroke from user. It doesn't have the 'expiration time' parameter though, but to implement the expiration logic you must have smth like an event loop, which, for my money, doesn't worth it most of the time. Such kind of loop is started when you create named windows, so waitKey was made to work with it. You could start the loop without any windows, just to support waitKey, but again - I personally don't think it's a good design decision.
2) I believe this API is a legacy from the 'only C' days of OpenCV. It looks kind of awkward today and have potential drawbacks, but since highgui module is mostly meant for debugging, I believe it's good enough to have the work done.
That's just my opinion, I hope it makes some sense to you, too.
- Assignee set to Alexander Smorkalov
- Status changed from New to Open
```
##### Steven Puttemans on 2013-08-14 19:48
```
I can find myself in your opinion and yes there are probably other ways around this.
Thanks for your input!
Just a remark about the legacy style. I think if we want people to be pushed towards the newer C++ interface, even if it is for debugging, their should be a uniform use of objects and their functionality. That is why I think this suggestion is still valid :)
```
| auto-transferred,priority: normal,feature,category: highgui-gui | low | Critical |
97,427,029 | opencv | support for INT-Mat for remapBilinear OR adding note to documentation | Transferred from http://code.opencv.org/issues/3220
```
|| Thomas Zeugner on 2013-08-14 10:52
|| Priority: Normal
|| Affected: None
|| Category: imgproc, video
|| Tracker: Feature
|| Difficulty: Easy
|| PR:
|| Platform: None / None
```
## support for INT-Mat for remapBilinear OR adding note to documentation
```
Hallo there,
the function "remapBilinear" (modules/imgproc/src/imgwarp.cpp : 2465) does not support INT as element for Mat-Type so the call of warpPerspective fails with a crash.
Maybe it is only the "linear_tab[]" (modules/imgproc/src/imgwarp.cpp : 3098) Array that has to be fixed??
Example code:
<pre>
#include "opencv/cv.h"
#include "opencv/highgui.h"
int main( int argc, char** argv )
{
src = Mat::eye(480 , 640 , CV_32S);
cv::Mat dest;
Mat transMat = (Mat_<double>(3,3) << 0.83665, -0.0946, 35.96591, 0.00622, 0.49884, 92.2776, 0, -0.00029, 0.94627);
cv::Size size( src.cols, src.rows);
warpPerspective (src, dest, transMat, size, WARP_INVERSE_MAP | INTER_LINEAR ); //- OR: INTER_CUBIC | INTER_LINEAR
imshow( "dest", dest );
waitKey(0);
}
</pre>
If it is not possible to add this functionality then PLEASE add a note to the documentation of (warpPerspective, remap, warpAffine, resize, ...) . otherwise this is pretty confusing ;-)
Thank you very much!
Note: I work with a thermographic camera that returns one channel values as mK, so I have a value range of about 0 to 2 000 000 per pixel. Of cause, I could convert this values to float but I'm worried about the performance.
```
## History
##### Sergei Nosov on 2013-08-14 14:53
```
Hi, Thomas!
I'm afraid there'll be some issues with your solution. But you're welcome to try it out, and submit a pull request with the fix and corresponding tests to https://github.com/Itseez/opencv . Alternatively, you can submit a PR on docs.
P.S. I would suggest to try converting to float (or double, if you'll encounter accuracy problems) and measuring. I guess it won't be much (if any) slower.
```
##### Sergei Nosov on 2013-08-14 15:03
```
- Status changed from New to Open
```
##### Vadim Pisarevsky on 2015-05-13 16:01
```
looks like it's long overdue with the reply. On most platforms, even low-power ARM's, there is now hardware support for single-precision floating-point and for this operation (warpAffine/warpPerspective/remap) floating-point is needed anyway, so the suggestion is to convert data to floating-point - performance will be better. Integers till 2**24 ~ 10**7 can be represented exactly in float's.
- Tracker changed from Bug to Feature
```
| auto-transferred,priority: normal,feature,category: imgproc,category: video | low | Critical |
97,427,096 | opencv | BFMatcher Indexing not suited for large scale | Transferred from http://code.opencv.org/issues/3221
```
|| Dominik Van Opdenbosch on 2013-08-14 12:09
|| Priority: Low
|| Affected: 2.4.0 - 2.4.6
|| Category: features2d
|| Tracker: Bug
|| Difficulty: Easy
|| PR:
|| Platform: x64 / Linux
```
## BFMatcher Indexing not suited for large scale
```
The way the indexing is done in the Brute Force Matcher Implementation is not suited for large scale scenarios.
The indexing is limited by the use of 32bit signed integer instead of (unsigned) 64 bit integer where available (long int):
modules/features2d/src/matchers.cpp
CV_Assert( (int64)imgCount*IMGIDX_ONE < INT_MAX )
```
## History
##### Sergei Nosov on 2013-08-14 15:00
```
Hi, Dominik!
I'm not sure it's clear what you actually want to say. Could you, please, be more specific?
- Target version set to 2.4.7
- Status changed from New to Incomplete
```
##### Mathieu Barnachon on 2013-08-22 03:29
```
Did you mean: replace INT_MAX by INT64_MAX?
But update variable must be a int64 also.
Is it suitable for OpenCV to use 64 bits variables (for multi-plateform compatibility I mean)?
```
##### Alexander Smorkalov on 2013-11-28 05:02
```
- Affected version changed from 2.4.6 (latest release) to 2.4.0 -
2.4.6
- Target version changed from 2.4.7 to 2.4.8
```
##### Alexander Smorkalov on 2013-12-30 10:38
```
- Target version changed from 2.4.8 to 2.4.9
```
##### Alexander Smorkalov on 2014-04-30 19:05
```
- Target version changed from 2.4.9 to 2.4.10
```
| bug,auto-transferred,priority: low,affected: 2.4,category: features2d | low | Critical |
97,427,130 | opencv | libc++ build no longer works due to use of stdlib in gpulegacy | Transferred from http://code.opencv.org/issues/3228
```
|| Miroslav Kobetski on 2013-08-21 12:56
|| Priority: Normal
|| Affected: branch 'master' (2.9)
|| Category: gpu (cuda)
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Mac OSX
```
## libc++ build no longer works due to use of stdlib in gpulegacy
```
Compiling opencv master (2.9) doesn't work with clang and libc++ on Mac. (didn't set the std=c++11 flag as that gives rise to a new set of errors).
gpulegacy fails during link time, and gpu cannot be compiled without gpulegacy (due to modules/gpuwarping/src/pyramids.cpp:235:12: error: use of undeclared identifier 'numLayers')
Culprits (found so far): iostream and vector in NCVBroxOpticalFlow and NCV cause link errors:
"std::ios_base::Init::Init()", referenced from:
__static_initialization_and_destruction_0(int, int) in cuda_compile_generated_NCV.cu.o
__static_initialization_and_destruction_0(int, int) in cuda_compile_generated_NCVBroxOpticalFlow.cu.o
"std::ios_base::Init::~Init()", referenced from:
___tcf_0 in cuda_compile_generated_NCV.cu.o
___tcf_0 in cuda_compile_generated_NCVBroxOpticalFlow.cu.o
"std::__throw_length_error(char const*)", referenced from:
```
## History
##### Alexander Shishkov on 2013-08-27 08:37
```
- Assignee set to Anatoly Baksheev
- Category set to gpu (cuda)
```
##### Alexander Shishkov on 2013-08-27 08:38
```
Hello Miroslav! Thx for reporting!
Maybe you know some solution for this problem?
```
##### Alexander Shishkov on 2013-08-28 19:52
```
- Status changed from New to Open
```
##### Miroslav Kobetski on 2013-11-07 10:25
```
Sorry for the late answer, just noticed that this is still open. As an update: The original problem with gpulegacy was due to an incorrect naming in the function params. The input was "nLayers", but it expected "numLayers". This particular issue seems fixed in the latest master.
I will try to get around to looking at the rest of this problem in a week or two, and can then post an update if I can reproduce it and if I have a solution
```
| bug,auto-transferred,priority: normal,affected: 3.4,category: gpu/cuda (contrib) | low | Critical |
97,427,194 | opencv | Add a way to detect circles grid with unknown size | Transferred from http://code.opencv.org/issues/3237
```
|| Nikolay Tasev on 2013-08-26 11:00
|| Priority: Normal
|| Affected: None
|| Category: calibration, 3d
|| Tracker: Feature
|| Difficulty: Medium
|| PR:
|| Platform: None / None
```
## Add a way to detect circles grid with unknown size
```
There should be a way to use findCirclesGrid() without knowing the pattern size.
This will be useful to detect or at least give a proposition for the user what is
the size of an unknown grid from a single frame. After the size is determined, the
calibration can proceed as before on multiple frames.
I have looked in the latest source code for the symmetrical circle grid. The
implementation first detects the grid and then checks if it matches the pattern
size. If it doesn't just rejects it. So for an unknown size this restriction
should be removed and the detected grid should be returned.
```
## History
##### Steven Puttemans on 2013-08-26 11:12
```
Feel free to submit a pull request with the proposed adaptations. This is the quickest way to see your request integrated into the OpenCV source code. More information on how to do so, take a look at : http://code.opencv.org/projects/opencv/wiki/How_to_contribute
```
##### Alexander Shishkov on 2013-08-27 08:47
```
Ilya, could you comment this issue?
- Difficulty changed from Easy to Medium
- Assignee set to Ilya Lysenkov
```
##### Alexander Shishkov on 2013-08-28 19:54
```
- Status changed from New to Open
```
##### Ilya Lysenkov on 2013-09-02 07:09
```
I agree that such a feature could be useful but the pattern size is used a lot internally in the detection algorithms. Yes, we have the final check of the pattern size but former detection will not work out of the box in case of unknown pattern size. So this is not so easy to implement with current algorithms.
```
| auto-transferred,priority: normal,feature,category: calib3d | low | Minor |
97,427,223 | opencv | Allow opencv_traincascade to load Haar wavelets from a file | Transferred from http://code.opencv.org/issues/3241
```
|| Ramiro Pereira de Magalhães on 2013-08-26 23:09
|| Priority: Normal
|| Affected: None
|| Category: objdetect
|| Tracker: Feature
|| Difficulty:
|| PR:
|| Platform: None / None
```
## Allow opencv_traincascade to load Haar wavelets from a file
```
Currently the opencv_traincascade application generates the Haar wavelets it will use for trainning. I think this application should be able to load the Haar wavelets configuration (amount and format of the rectangles and weights) from a file formated in a yet to be defined way.
The current Haar wavelet generation process could be extracted to another tool that generates a file with the default wavelets for use with the opencv_traincascade.
Thoughs?
```
## History
##### Alexander Shishkov on 2013-08-27 08:49
```
Hello Ramiro,
thanks for submitting this issue. Unfortunately we have limited resources at the moment. So if there is a chance if you have time to implement these features that would be highly appreciated! (Please see http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute for details).
Thank you in advance!
- Assignee set to Marina Kolpakova
```
##### Alexander Shishkov on 2013-08-27 08:49
```
Marina, could you please comment this issue?
```
##### Ramiro Pereira de Magalhães on 2013-08-27 18:44
```
I'm currently adapting the OpenCV code to do something like for an experiment I'm working on and I might be able to share it with the community. I'm just wondering if you have more specific guidelines or requirements about this proposed feature that I should consider.
```
##### Steven Puttemans on 2013-08-28 06:46
```
In addition to this, I still think it should be pretty nice if we could create an interface to automatically load other feature types into the traincascade algorithm. However I have no idea on how difficult this will actually be, but I am up for help if needed.
```
##### Alexander Shishkov on 2013-08-28 19:54
```
- Status changed from New to Open
```
##### Vadim Pisarevsky on 2015-05-28 16:57
```
sounds like a feature request
- Tracker changed from Bug to Feature
```
| auto-transferred,priority: normal,feature,category: objdetect | low | Critical |
97,427,279 | opencv | Update FLANN Library | Transferred from http://code.opencv.org/issues/3249
```
|| Dominik Van Opdenbosch on 2013-09-05 08:44
|| Priority: Low
|| Affected: None
|| Category: flann
|| Tracker: Feature
|| Difficulty: Easy
|| PR:
|| Platform: None / None
```
## Update FLANN Library
```
There are several issues with the FLANN Library which have been fixed in the flann master branch but not pulled
to the OpenCV Code.
https://github.com/mariusmuja/flann/commit/b72d526463e892bfc4eb30bf0c640a3cb5b9bd69#src/cpp/flann/algorithms/lsh_index.h
```
## History
##### Anna Kogan on 2013-09-05 08:50
```
Hello Dominik,
Thank you for reporting the problem. If you could make a pull request with those fixes, it would be very appreciated!
- Assignee deleted (Marius Muja)
- Status changed from New to Open
```
##### Alexander Smorkalov on 2013-09-09 10:01
```
- Target version set to 2.4.7
- Assignee set to Roman Donchenko
```
##### Prasan Samtani on 2013-09-09 17:09
```
I can take a look at this if nobody else is doing so.
It appears that the flann repo on github is significantly different from the module in opencv, is there a reason behind this?
```
##### Guanta Alecho on 2013-09-18 14:02
```
Prasan Samtani wrote:
> I can take a look at this if nobody else is doing so.
Cool, this would be great! I'd appreciate an updated version!
>
> It appears that the flann repo on github is significantly different from the module in opencv, is there a reason behind this?
Afaik it's just very outdated.
```
##### Prasan Samtani on 2013-09-19 14:44
```
Alright, I'll try and do so this weekend.
```
##### Alexander Smorkalov on 2013-11-27 10:51
```
@Roman Donchenko, please have a look. I believe, that this fix has been already cherry-picked to OpenCV.
- Affected version changed from branch 'master' (2.9) to 2.4.0 - 2.4.6
```
##### Alexander Smorkalov on 2013-11-28 05:02
```
- Target version changed from 2.4.7 to 2.4.8
```
##### Roman Donchenko on 2013-11-29 08:07
```
No, that one wasn't.
Honestly, our copy of FLANN is pretty much unmaintained; there are even plans to ditch it by 3.0. If you (well, anyone reading this) want to cherry-pick some critical fixes, send us a pull request and we could merge it, but a wholesale update is unlikely to happen.
- Priority changed from Normal to Low
```
##### Alexander Smorkalov on 2013-11-29 10:58
```
- Target version changed from 2.4.8 to Next Hackathon
```
##### Maksim Shabunin on 2015-07-21 09:08
```
- Affected version changed from 2.4.0 - 2.4.6 to branch 'master'
(3.0-dev)
```
##### Maksim Shabunin on 2015-07-21 09:08
```
- Tracker changed from Patch to Feature
```
| auto-transferred,feature,priority: low,category: flann | low | Minor |
97,427,320 | opencv | cv2.EM.getMeans() does not work | Transferred from http://code.opencv.org/issues/3257
```
|| Devendra Goyal on 2013-09-10 19:38
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.5
|| Category: python bindings
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x86 / Mac OSX
```
## cv2.EM.getMeans() does not work
```
trying to access the means after applying EM on a GMM:
using code detailed here: http://docs.opencv.org/2.3/modules/ml/doc/expectation_maximization.html
but I get this error...I am using opencv 2.4.5
<pre>
AttributeError: 'cv2.EM' object has no attribute 'getMeans'
</pre>
How do I access the means and covariances using Python?!
```
## History
##### Alexander Smorkalov on 2013-09-11 06:02
```
Hello
You use documentation from older OpenCV version then you use. The docs for 2.4.5 is here: http://docs.opencv.org/2.4.5/modules/ml/doc/expectation_maximization.html#cvem-getmeans. it looks like wrapper for this class was removed after code refactoring.
Vadim, please comment the issue.
- Priority changed from Blocker to Normal
- Assignee set to Vadim Pisarevsky
- Category set to python bindings
```
##### Anna Kogan on 2013-10-03 07:56
```
- Description changed from trying to access the means after applying
EM on a GMM: using code detailed he... to trying to access the means
after applying EM on a GMM: using code detailed he... More
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,category: python bindings,affected: 2.4 | low | Critical |
97,427,357 | opencv | OpenCV with_Openexr build errors | Transferred from http://code.opencv.org/issues/3259
```
|| git gat on 2013-09-11 08:34
|| Priority: Normal
|| Affected: 2.4.6 (latest release)
|| Category: highgui-images
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Windows
```
## OpenCV with_Openexr build errors
```
Hi,
I get many build errors when compiling the latest OpenCV with openexr.
when I don't use with_openexr all is good!
Here is the error:
<pre>
Error 156 error: "restrict" has already been declared in the current scope \3rdparty\openexr\Imath\ImathMatrix.h 2530
Error 158 error: "restrict" has already been declared in the current scope \3rdparty\openexr\Imath\ImathMatrix.h 2531
Error 155 error: expected a ";" \3rdparty\openexr\Imath\ImathMatrix.h 2529
Error 157 error: expected a ";" \3rdparty\openexr\Imath\ImathMatrix.h 2530
Error 159 error: expected a ";" \3rdparty\openexr\Imath\ImathMatrix.h 2531
Error 160 error: identifier "ap" is undefined \3rdparty\openexr\Imath\ImathMatrix.h 2535
Error 162 error: identifier "bp" is undefined \3rdparty\openexr\Imath\ImathMatrix.h 2540
Error 161 error: identifier "cp" is undefined \3rdparty\openexr\Imath\ImathMatrix.h 2540
</pre>
I use Windows 7, VS 2008.
```
## History
##### Alexander Smorkalov on 2013-09-11 10:22
```
Hello git gat,
Could you provide some code sample to reproduce build issue?
- Assignee set to Alexander Smorkalov
- Category set to highgui-images
```
##### Alexander Smorkalov on 2013-09-12 10:52
```
- Target version set to Next Hackathon
```
##### Anna Kogan on 2013-10-03 07:58
```
- Affected version changed from branch '2.4' to 2.4.6 (latest release)
- Description changed from Hi, I get many build errors when compiling
the latest OpenCV with openexr.... to Hi, I get many build errors
when compiling the latest OpenCV with openexr.... More
- Status changed from New to Open
```
##### Alexander Smorkalov on 2013-10-31 08:46
```
I cannot reproduce build issue with Visual Studio 2008 and Visual Studio 2012. If you look at 3rdparty/openexr/Imath/ImathPlatform.h IMATH_RESTRICT macro is defined empty for MSVC. It means that you use some other compiler like Intel C++ compiler or you have some other version of OpenEXR lib in include path and compiler uses mix of headers from different locations. Unfortunately, I do not have ICC and cannot check compilation with it. I close the ticket.
- Status changed from Open to Cancelled
```
##### Tim Zuercher on 2014-07-27 11:05
```
I was experiencing this same issue using the ICC compiler. Reassigning the IMATH_RESTRICT macro from "restrict" to empty fixes the issue.
There are also a overwhelming number of warnings for signed/unsigned integer comparisons.
- Status changed from Cancelled to Incomplete
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: imgcodecs | low | Critical |
97,427,499 | opencv | MSER detecting non-extremal regions | Transferred from http://code.opencv.org/issues/3272
```
|| Petra Bosilj on 2013-09-19 10:02
|| Priority: Normal
|| Affected: pre 2.4 (deprecated)
|| Category: features2d
|| Tracker: Bug
|| Difficulty: Hard
|| PR:
|| Platform: Any / Any
```
## MSER detecting non-extremal regions
```
I found that the MSER detector in v.2.3.1 is giving some regions which are not extremal -- there are border pixels of same intensity as some of the pixels inside the regions. I checked the code for the current version and it did not seem to be changed at all, and I did not find any bug report about this so I think the issue is still present.
I gave a detailed explanation, with parameters and the image I used, and some debug output here: http://answers.opencv.org/question/20926/mser-detection-error-detected-region-not-extremal/
```
## History
##### Anna Kogan on 2013-10-03 08:14
```
Hello Petra,
Thank you for reporting the issue. Did you try using new versions of the library? Is the bug reproduced there?
If you could figure out how to solve the problem, a "contribution":http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute would be very appreciated!
- Status changed from New to Open
```
##### Petra Bosilj on 2013-10-08 14:48
```
Hello,
sorry, did not try with the new version. But what I did was browse through the code in the new version, and code for MSER implementation does not seem to have any changes when compared to the old version.
I'm glad to report the bug and details about it, but unfortunately (and especially since all OpenCV code is optimized for speed and so hard to read), I do not think that I'll have time to try and understand and debug the code.
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: features2d | low | Critical |
97,427,577 | opencv | projectPoints and 3d points behind camera. | Transferred from http://code.opencv.org/issues/3281
```
|| Rafael Saracchini on 2013-09-23 11:42
|| Priority: Normal
|| Affected: None
|| Category: calibration, 3d
|| Tracker: Feature
|| Difficulty:
|| PR:
|| Platform: None / None
```
## projectPoints and 3d points behind camera.
```
When making some tests with the procedure "projectPoints", I noticed that it does not make distinction whether a point is behind the camera or not. This feature is very useful, for example, to draw a small 3D cloud onto the scene or to verify the extrinsic camera matrix of a calibration procedure is really valid. I believe that it could be done accepting a 4XN input and returning a 3XN output (that is, accept homogeneous coordinates as input and output) so that it does not break compatibility the original function calls.
All the best.
Rafael Saracchini
```
## History
##### Vladislav Vinogradov on 2013-09-24 08:06
```
Hello Rafael Saracchini!
Thank you for your proposal!
Unfortunately we have a limited human resources and we would be very appreciated if you have time to implement this feature by yourself and provide a patch to community.
Please see http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute for details.
- Assignee set to Rafael Saracchini
```
##### Vladislav Vinogradov on 2013-09-26 08:44
```
- Category set to calibration, 3d
```
##### Anna Kogan on 2013-10-03 08:59
```
- Status changed from New to Open
```
| auto-transferred,priority: normal,feature,category: calib3d | low | Minor |
97,427,736 | opencv | Face Detection training image problem | Transferred from http://code.opencv.org/issues/3300
```
|| Naval Kishore Gandhi on 2013-10-03 11:50
|| Priority: Low
|| Affected: 2.4.0 - 2.4.6
|| Category: java bindings
|| Tracker: Bug
|| Difficulty: Medium
|| PR:
|| Platform: x64 / Windows
```
## Face Detection training image problem
```
The FaceRecognizer class of java have a function named predict which is used to compare the facial image and a set of training images and then it predicts which face is this.
The bug is that the predict methods throws an exception if the training image's width is not a multiple of 8, I tried to get prediction work with training images of 125X150 resolution.
Predict method works if I use LBPH implementation of Facerecognizer, but if i use Fisherman or Eigenface it gave me this exception:-
<pre>
OpenCV Error: Image step is wrong (The matrix is not continuous, thus its number of rows can not be changed) in unknown function, file ..\..\..\src\opencv\modules\core\src\matrix.cpp, line 802
Exception in thread "pool-1-thread-4" java.lang.RuntimeException: ..\..\..\src\opencv\modules\core\src\matrix.cpp:802: error: (-13) The matrix is not continuous, thus its number of rows can not be changed.
</pre>
However, as a suggestion of [[http://code.google.com/p/javacv/issues/detail?id=368]] when I changed the resolution of training images to 128X150. it worked without any exception.
```
## History
##### Anna Kogan on 2013-10-04 07:34
```
Hello Naval,
Thank you for reporting the problem! If you could work on a solution for the issue, a "contribution":[[How_to_contribute]] would be very appreciated!
- Priority changed from Blocker to Low
- Description changed from The FaceRecognizer class of java have a
function named predict which is used ... to The FaceRecognizer class
of java have a function named predict which is used ... More
```
##### Anna Kogan on 2013-10-04 07:35
```
- Status changed from New to Open
```
##### Alexander Smorkalov on 2013-11-27 10:48
```
@Andrey Pavlenko, please have a look.
- Affected version changed from 2.4.6 (latest release) to 2.4.0 -
2.4.6
```
##### Alexander Smorkalov on 2013-11-28 06:33
```
- Target version changed from 2.4.7 to 2.4.8
```
##### Alexander Smorkalov on 2013-12-30 10:38
```
- Target version changed from 2.4.8 to 2.4.9
```
##### Alexander Smorkalov on 2014-04-30 19:05
```
- Target version changed from 2.4.9 to 2.4.10
```
| bug,auto-transferred,priority: low,affected: 2.4,category: java bindings | low | Critical |
97,427,779 | opencv | getAffineTransform(): No error handling with aligned points | Transferred from http://code.opencv.org/issues/3305
```
|| Arnaud Ramey on 2013-10-07 11:59
|| Priority: Normal
|| Affected: 2.4.6 (latest release)
|| Category: imgproc, video
|| Tracker: Bug
|| Difficulty: Medium
|| PR:
|| Platform: Any / Any
```
## getAffineTransform(): No error handling with aligned points
```
Let A and B be sets of 3 points.
If the points in A or in B are aligned,
no affine transformation exists that match A to B in most cases.
However, in such a case, @cv::getAffineTransform()@ returns a matrix with no complain
but containing invalid data.
It would be nice to cast an error, or at least to specify the behavior in the documentation.
Minimal example:
<pre>
cv::Point2f A_sample[3];
A_sample[0] = cv::Point2f(8, 9);
A_sample[1] = cv::Point2f(40, 41);
A_sample[2] = cv::Point2f(47, 48);
cv::Point2f B_sample[3];
B_sample[0] = cv::Point2f(7.37465, 11.8295);
B_sample[1] = cv::Point2f(15.0113, 12.8994);
B_sample[2] = cv::Point2f(38.9943, 9.56297);
cv::Mat trans = cv::getAffineTransform(A_sample, B_sample);
std::cout << "trans:" << trans << std::endl;
___________________
Output:
___________________
trans:[-8245045546463036, 8245045546463037, -8245045546463042;
1319381983356685, -1319381983356685, 1319381983356699]
</pre>
```
## History
##### Kirill Kornyakov on 2013-10-08 07:52
```
Arnaud Ramey, thank you for the report, but it would be great if you could work on a pull request: http://code.opencv.org/projects/opencv/wiki/How_to_contribute.
- Difficulty changed from Easy to Medium
- Operating System changed from Linux to Any
- HW Platform changed from x86 to Any
- Target version set to Next Hackathon
- Description changed from Let A and B be sets of 3 points. If the
points in A or in B are aligned, no ... to Let A and B be sets of 3
points. If the points in A or in B are aligned, no ... More
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,category: imgproc,category: video,affected: 2.4 | low | Critical |
97,427,842 | opencv | svm: different results between iOS simulator & iPhone | Transferred from http://code.opencv.org/issues/3308
```
|| Boris Mansencal on 2013-10-08 14:58
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.6
|| Category: ios
|| Tracker: Bug
|| Difficulty: Medium
|| PR:
|| Platform: ARM / iOS
```
## svm: different results between iOS simulator & iPhone
```
Hello,
It seems there is a bug when SVM classifier is executed on a real iPhone. We do not get the same results than when it is executed on iOS simulator (or MacOSX or Linux).
The attached program should show the bug.
The attached program load two files : an SVM model and a data file [also attached to this bug].
The data file contains both data and results of classification for this data with this SVM model (computed on linux).
The program load the model, load the data, classify the data with the model and check if the results are the same than those expected (loaded from file).
Once compiled, the program is launched liked this :
./test data.yml model.xml
The program executed on linux, mac, or iOS simulator gives the same results (as those obtained on linux).
But when executed on a real iPhone, the results are different than those expected.
We use the same version of OpenCV on all platforms : 2.4.6
Linux platform is : Intel Xeon, Fedora 19, g++ 4.8.1
MacOSX platform is : Intel, MacOSX 10.8.5, Xcode 5 (5A1413) + associated iOS simulator
iPhone platform is iPhone 5 with iOS 7, or iPhone 5S with iOS 7.0.2
We also tested on a MacBookPro, Intel Core2Duo, MacOSX 10.7.4, Xcode 4.6.3, with OpenCV 2.4.3 and we get the expected results.
On a real iPhone, we tested 2.4.6 (downloaded framework or compiled from source) & current git version, and in all cases we get different results than under the simulator.
Let me know if you need more details or to test certain things on these configurations.
Boris.
```
## History
##### Kirill Kornyakov on 2013-10-09 15:44
```
Thank you for this great report, but please do not expect fast response... Nobody is maintaining the iOS port at the moment, so if you really need the fix soon (and send the pull request), I would suggest to debug the issue yourself or workaround it somehow.
- Difficulty set to Medium
- Status changed from New to Open
```
##### Alexander Smorkalov on 2013-11-27 10:44
```
@Alexander Shishkov, any comments from your side?
- Affected version changed from 2.4.6 (latest release) to 2.4.0 -
2.4.6
```
##### Alexander Smorkalov on 2013-11-28 06:32
```
- Target version changed from 2.4.7 to 2.4.8
```
##### Alexander Smorkalov on 2013-12-30 10:38
```
- Target version changed from 2.4.8 to 2.4.9
```
##### Adam Lang on 2014-03-06 21:31
```
I have the same problem, SVM classification works in Simulator but not on the iPhone. On iPhone i get always the max float value as the distance result. Is there any solution yet or any news?
```
##### Andreas Hörsken on 2014-04-24 12:09
```
I tested the sample above with an iPad 4 and iPad Air (x64) under both, iOS 7 and iOS 7.1 and got no errors (wether with the simulator nor the device). Is it solved or does it just not work with the iphone? Did you test it with a current device?
```
##### Alexander Smorkalov on 2014-04-30 19:05
```
- Target version changed from 2.4.9 to 2.4.10
```
| bug,auto-transferred,priority: normal,affected: 2.4,platform: ios/osx | low | Critical |
97,428,101 | opencv | Problem with retrieving frames from h264 video [Linux Ubuntu versus Windows 7] | Transferred from http://code.opencv.org/issues/3326
```
|| Steven Puttemans on 2013-10-17 08:37
|| Priority: Normal
|| Affected: 2.4.6 (latest release)
|| Category: highgui-video
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Windows
```
## Problem with retrieving frames from h264 video [Linux Ubuntu versus Windows 7]
```
I have a video and a file containing xml annotations.
The video file is an infra red video feed of a car park, where I want to perform detection of persons and video.
We have a problem with the reading of the actual video file and matching the annotations.
ON WINDOWS 7 X64 OPENCV 2.4.6
When running the code supplied in attachements called 'inlezen.cpp', errors get outputted, connected to the h264 encoding. The result is that the bounding box of the first frame (green and red box overlap but ignore that part) gets printed at the extracted location that was matched with the first annotation in the file. If you look close, you can see that the actual persons is in front of the annotation.
'!error_image.PNG!'
ON LINUX UBUNTU 12.06 OPENCV 2.4.6
However, running the same code on linux, only produces the first 2 error lines as seen in the image, not the others. Result is that the bounding box matches perfectly with the person intended to encapsulate.
I am afraid this has something to do with the video itself and not with openCV, but I want to make sure nevertheless. The video can be found at this link: http://www.eavise.be/tobcat/opencv/video.avi
The annotation file is in the extras added to this post, but video was to large.
The test archive contains all functionality needed to extract the xml from the xgtf file and the inlezen.cpp file contains all the functionality used to read in the frames. You can see that there is some logic to match the correct annotations with the correct frame, but that works correctly (like illustrated on Linux).
I am just wondering why there is a difference between both systems, VideoCapture uses the same ffmpeg library right?
```
## History
##### Steven Puttemans on 2013-10-17 08:39
```
Also as an addition, the number that is printed in the output, is a check that the current frame 1157 is the same as the first annotation that contains a human, which is class 11 in the xml file.
```
##### Maria Dimashova on 2013-10-17 09:11
```
- Assignee set to Vadim Pisarevsky
```
##### Steven Puttemans on 2013-10-18 12:42
```
I can add that I have tested it in x64 and x86 of the operating systems and that the conflict stays the same.
It thus must have something to do with the h264 encoding of the ffmpeg library i think...
```
##### Dmitry Retinskiy on 2013-10-28 17:02
```
- Target version changed from 2.4.7 to 2.4.8
- Status changed from New to Open
```
##### Steven Puttemans on 2013-10-30 09:25
```
I have just pulled the latest 2.4 version from GIT, still no changes, not even with the new ffmpeg library included.
Is the latest ffmpeg library provided? Of is there a possibility of building with an even more recent version?
```
##### Alexander Smorkalov on 2013-12-30 10:38
```
- Target version changed from 2.4.8 to 2.4.9
```
##### Alexander Smorkalov on 2014-04-30 19:05
```
- Target version changed from 2.4.9 to 2.4.10
```
| bug,auto-transferred,priority: normal,category: videoio,affected: 2.4 | low | Critical |
97,428,167 | opencv | CommandLineParser | Transferred from http://code.opencv.org/issues/3330
```
|| Ilya Lavrenov on 2013-10-21 11:42
|| Priority: Normal
|| Affected: branch '2.4' (2.4-dev)
|| Category: core
|| Tracker: Bug
|| Difficulty: Medium
|| PR:
|| Platform: Any / Any
```
## CommandLineParser
```
cv::CommandLineParser gives segmentation fault when:
- we pass arguments like '-', '--', '---' and so on (see https://github.com/Itseez/opencv/pull/1590)
- we pass arguments as --argument_name, but inside the program try to get value like this:
@parser.get<int>("argument_name");@
```
## History
##### Ilya Lavrenov on 2013-10-21 12:33
```
also, cmd parser fails in constructor with keys
@"{ h | help | | show available options }"@
```
| bug,auto-transferred,priority: normal,category: core,affected: 2.4 | low | Critical |
97,428,379 | opencv | Regarding imshow (OpenCV with OpenGL support) | Transferred from http://code.opencv.org/issues/3352
```
|| Philipp Werner on 2013-10-30 13:04
|| Priority: Normal
|| Affected: 2.4.6 (latest release)
|| Category: highgui-gui
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Windows
```
## Regarding imshow (OpenCV with OpenGL support)
```
The following code is giving flickering window without displaying any image.
<pre>
Mat img = imread("image.jpg",CV_LOAD_IMAGE_ANYCOLOR);
namedWindow("window", WINDOW_OPENGL);
imshow("window",img);
waitKey(0);
</pre>
and following code is working fine.
<pre>
Mat img = imread("image.jpg",CV_LOAD_IMAGE_ANYCOLOR);
namedWindow("window", WINDOW_AUTOSIZE);
imshow("window",img);
waitKey(0);
</pre>
Further, imshow() has the mentioned faulty behaviour if namedWindow() is not called explicitly and OpenCV has been compiled with OpenGL support.
I already found that adding a call to waitKey() before imshow() solves the problem, probably because it handles messages that need to be handled in between the creation of the window and the displaying of an image. So, this works:
<pre>
Mat img = imread("image.jpg",CV_LOAD_IMAGE_ANYCOLOR);
namedWindow("window", WINDOW_OPENGL);
waitKey(0);
imshow("window",img);
waitKey(0);
</pre>
However, this solution is neither documented nor consistent with OpenCV versions compiled without OpenGL support.
I am using Windows 7 64 bit, OpenCV 2.4.6.0 built with default settings + OpenGL-support. The same behaviour has been reported for OpenCV 2.4.3 with Windows 7 32 bit:
http://answers.opencv.org/question/7916/regarding-imshow-opencv-compiled-with-opengl/
```
## History
##### Dmitry Retinskiy on 2013-11-01 05:44
```
Hi Philipp,
could you try to compile and run .../samples/gpu/opengl.cpp and reply if you observe the same behavior?
```
##### Dinar Ahmatnurov on 2013-11-08 08:26
```
Hi Philipp,
Any updates?
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,category: highgui-gui,affected: 2.4 | low | Critical |
97,428,609 | opencv | Vec Matx like initializers | Transferred from http://code.opencv.org/issues/3380
```
|| José Franco on 2013-11-12 22:59
|| Priority: Normal
|| Affected: None
|| Category: core
|| Tracker: Feature
|| Difficulty: Easy
|| PR:
|| Platform: None / None
```
## Vec Matx like initializers
```
Hello,
Despite cv::Vec<_Tp, int> being equivalent to cv::Matx<_Tp, int, 1>, it is currently not possible to initialize them with the Matx<_Tp, int, int> initializers cv::Matx<_Tp, int, int>::zeros(), ones(), etc. It is also not clear how to assign a cv::Matx<_Tp, int, 1> to a cv::Vec<_Tp, int>. These initializers would be very helpful.
Thanks
```
## History
##### Ivan Korolev on 2013-11-13 08:38
```
Hi José,
thanks for feature request.
Yes, actually these initializers would be helpful. Unfortunately, now we have limited human resources. So if you could implement this functionality and send a pull request on GitHub (http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute), it would be much appreciated!
- Status changed from New to Open
```
##### Rohit Girdhar on 2014-01-29 13:32
```
Hi Ivan,
I'm a student and just getting started with contributing to openCV (though have a lot of experience in using openCV).
I've already forked, cloned, built and played a bit with the source code.
I want to work on this task, and think it'd be better to run my ideas through you before starting to code.
If I understand the question correctly, we essentially want the ::zeros() kind-of initializers for the Vec.
In my understanding, Vec is implemented using Mat in modules/core/include/opencv2/core/matx.hpp (line 869).
So essentially, I need to add another function there for initialization.
Please let me know if I'm on the right track or not.
I'm really a noob in working on open source, so please forgive me for mistakes.
Thanks a lot!
```
##### Rohit Girdhar on 2014-02-09 12:51
```
PR for ::zeros() initializer (https://github.com/Itseez/opencv/pull/2307)
```
#2307
| auto-transferred,priority: normal,feature,category: core | low | Minor |
97,428,652 | opencv | Bug in RotatedRect::boundingRect() method | Transferred from http://code.opencv.org/issues/3396
```
|| Hamed Habibi Aghdam on 2013-11-25 14:15
|| Priority: Low
|| Affected: None
|| Category: core
|| Tracker: Feature
|| Difficulty: Easy
|| PR:
|| Platform: None / None
```
## Bug in RotatedRect::boundingRect() method
```
I tried to get the smallest bounding box for a rotated rectangle. However, the result of RotatedRect::boundingRect() method is not correct all time and in some cases it return rectangles much bigger than the original one.
I checked the source code for of this method. There is a mathematical solution to this problem. I implemented and tested it. Given the major and minor axis as well as the amount of the rotation of the RotatedRect, the smallest bounding box can be calculated as follow:
///////////////////////////////////////////////////////////////////
float t1 = atan(-(majorAxe*sin(degree))/(minorAxe*cos(degree)));
float t2 = t1 + 3.1415;
float maxX = x + majorAxe*cos(t1)*cos(degree)-minorAxe*sin(t1)*sin(degree);
float minX = x + majorAxe*cos(t2)*cos(degree)-minorAxe*sin(t2)*sin(degree);
t1 = atan((minorAxe*cos(degree))/(majorAxe*sin(degree)));
t2 = t1 + 3.1415;
float maxY = y + minorAxe*sin(t1)*cos(degree)+majorAxe*cos(t1)*sin(degree);
float minY = y + minorAxe*sin(t2)*cos(degree)+majorAxe*cos(t2)*sin(degree);
if (minY > maxY)
{
float temp = minY;
minY = maxY;
maxY = temp;
}
if (minX > maxX)
{
float temp = minX;
minX = maxX;
maxX = temp;
}
Rect r(minX,minY,maxX-minX,maxY-minY);
///////////////////////////////////////////////////////
*To speed up the computations, the code can be optimized to avoid redundant sin and cos calls. Also, the code can be more optimized by utilizing simple Trigonometric equations.* So the more optimized code can be as follows:
float c_degree = cos(degree);
float s_degree = sin(degree);
float t1 = atan(-(majorAxe*s_degree)/(minorAxe*c_degree));
float c_t1 = cos(t1);
float s_t1 = sin(t1);
float w1 = majorAxe*c_t1*c_degree;
float w2 = minorAxe*s_t1*s_degree;
float maxX = x + w1-w2;
float minX = x - w1+w2;
t1 = atan((minorAxe*c_degree)/(majorAxe*s_degree));
c_t1 = cos(t1);
s_t1 = sin(t1);
w1 = minorAxe*s_t1*c_degree;
w2 = majorAxe*c_t1*s_degree;
float maxY = y + w1+w2;
float minY = y - w1-w2;
if (minY > maxY)
{
float temp = minY;
minY = maxY;
maxY = temp;
}
if (minX > maxX)
{
float temp = minX;
minX = maxX;
maxX = temp;
}
Rect r(minX,minY,maxX-minX,maxY-minY);
```
## History
##### Hamed Habibi Aghdam on 2013-11-25 16:10
```
I changed the opencv code and tested it. it works:
Rect RotatedRect::boundingRect() const
{
float degree = angle*3.1415/180;
float majorAxe = this->size.width/2;
float minorAxe = this->size.height/2;
float x = this->center.x;
float y =this->center.y;
float c_degree = cos(degree);
float s_degree = sin(degree);
float t1 = atan(-(majorAxe*s_degree)/(minorAxe*c_degree));
float c_t1 = cos(t1);
float s_t1 = sin(t1);
float w1 = majorAxe*c_t1*c_degree;
float w2 = minorAxe*s_t1*s_degree;
float maxX = x + w1-w2;
float minX = x - w1+w2;
/*#ifdef _DEBUG
printf("major=%f minor=%f x=%f y=%f deg=%f",majorAxe,minorAxe,x,y,degree);
#endif*/
t1 = atan((minorAxe*c_degree)/(majorAxe*s_degree));
c_t1 = cos(t1);
s_t1 = sin(t1);
w1 = minorAxe*s_t1*c_degree;
w2 = majorAxe*c_t1*s_degree;
float maxY = y + w1+w2;
float minY = y - w1-w2;
if (minY > maxY)
{
float temp = minY;
minY = maxY;
maxY = temp;
}
if (minX > maxX)
{
float temp = minX;
minX = maxX;
maxX = temp;
}
Rect r(minX,minY,maxX-minX+1,maxY-minY+1);
return r;
}
```
##### Steven Puttemans on 2013-11-25 18:30
```
Are you up for submitting a pull request? It will help the adaptation of the repository alot! Some hints on how to do so can be found here: code.opencv.org/projects/opencv/wiki/How_to_contribute! By submitting a pull request, the review process becomes more easier and build problems on different operating systems can be tested.
If you are not up for it, then give a shout so that someone can pick up your work and submit a pull.
- % Done changed from 0 to 50
- Target version set to 2.4.8
- Assignee set to Hamed Habibi Aghdam
- Status changed from New to Open
```
##### Daniil Osokin on 2013-12-03 07:57
```
- Category set to core
```
##### Alexander Smorkalov on 2013-12-30 10:38
```
- Target version changed from 2.4.8 to 2.4.9
```
##### Daniil Osokin on 2014-01-11 08:24
```
Hamed, please, attach a test sample with wrong bounding rect, otherwise this ticket will be closed.
- Affected version changed from 2.4.7 (latest release) to 2.4.8
(latest release)
```
##### Hamed Habibi Aghdam on 2014-01-16 16:10
```
Daniil,
I just saw your message. I will attach a picture of the wrong bounding rect in the next 4 days.
Regards,
```
##### Hamed Habibi Aghdam on 2014-01-16 16:57
```
Using following:
Mat mm = Mat::zeros(220,220,CV_8UC1);
RotatedRect rect2(Point(101.33,109.84),Size2f(177.04,109.25),73.71);
Rect r = rect2.boundingRect();
ellipse(mm,rect2,Scalar(255,255,255),2);
rectangle(mm,r,Scalar(255,255,255),2);
imshow("test",mm);
waitKey(0);
We get the following result:
!http://oi44.tinypic.com/27xjvxe.jpg!
However, after I changed the source code to the one that I mentioned before we can get the following result:
!http://i39.tinypic.com/2d7zypv.jpg!
You can also test the previous code and the modified code on following RotatedRect:
RotatedRect rect1(Point(144.52,89.47),Size2f(186.35,138.56),68.28);
```
##### Alexander Smorkalov on 2014-04-30 19:05
```
- Target version changed from 2.4.9 to 2.4.10
```
##### Vadim Pisarevsky on 2015-05-01 20:22
```
RotatedRect::boundingBox is (as the name implies) finds bounding box for a rotated rectangle, not for a rotated ellipse. The proposed algorithm should be put into a separate function. I change it from bug to a feature request.
- Priority changed from Normal to Low
- Assignee deleted (Hamed Habibi Aghdam)
- Tracker changed from Bug to Feature
```
| auto-transferred,feature,priority: low,category: core | low | Critical |
97,428,685 | opencv | ByteBuffer support in Java | Transferred from http://code.opencv.org/issues/3405
```
|| Jose Gómez on 2013-11-28 03:02
|| Priority: Normal
|| Affected: None
|| Category: java bindings
|| Tracker: Feature
|| Difficulty:
|| PR:
|| Platform: None / None
```
## ByteBuffer support in Java
```
This is a feature request based on Q&A http://answers.opencv.org/question/23752/modifying-pixels-efficiently-in-java/
While in C/C++ one can directly access and modify the data of a Mat, currently, the only way of modifying the pixels in a Mat from Java is either accessing pixel by pixel using Mat.get and Mat.put, or copying the Mat to a temporary array, and after the change, another copy back to the Mat:
> http://answers.opencv.org/question/5/how-to-get-and-modify-the-pixel-of-mat-in-java/
However, this is very inefficient.
The JavaCV wrapper offers the possibility of creating a ByteBuffer pointing to the actual data, thus avoiding copies of data from and to the Mat (IplImage in case of JavaCV):
> http://stackoverflow.com/questions/9920449/converting-opencv-matrix-looping-to-javacv
> http://stackoverflow.com/questions/14877383/iplimage-pixel-access-javacv
Attached a performance using 4 different implementations of the same multi-step algorithm:
* "JavaCV - Gradient": Normal algorithm using the JavaCV wrapper, and accessing the data in the Mats using ByteBuffers
* "OpenCV - Gradientgetput": The same algorithm in the OpenCV wrapper, using Mat.get and Mat.put per each pixel
* "OpenCV - Gradient1array": The same algorithm in the OpenCV wrapper, copying the data from and to a Mat to an array for the most complex step.
* "OpenCV - Gradient": The same algorithm in the OpenCV wrapper, copying the data from and to a Mat to an array for all the steps that require custom proc# essing.
From those tests, we can observe that the most optimised OpenCV version, using arrays, is about 10%-15% slower than the JavaCV version using ByteBuffer. Furthermore, this version gets messy when it is required to mix OpenCV functions (which operate on Mats) and custom functions (which due to the performance has to operate on arrays).
```
## History
##### Nikita Manovich on 2013-12-11 08:33
```
Hi,
Thank you for the feature request. Could you please help us and prepare a pull request for the functionality? It will help us a lot. Here you can find all necessary information: http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute
- Assignee changed from Andrey Pavlenko to Jose Gómez
- Status changed from New to Open
```
| auto-transferred,priority: normal,feature,category: java bindings | low | Major |
97,428,760 | opencv | gpu::buildWarpPerspectiveMaps is not thread safe | Transferred from http://code.opencv.org/issues/3419
```
|| Todd Pepper on 2013-12-04 00:38
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.6
|| Category: gpu (cuda)
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: Any / Linux
```
## gpu::buildWarpPerspectiveMaps is not thread safe
```
It looks like gpu::buildWarpPerspectiveMaps is not thread safe due to the use of the float array c_warpMat below. I first noticed this when debugging an issue in porting my tile warping application from CPU to GPU. The gpu::warpPerspective and gpu::buildWarpPerspectiveMaps calls require mutex locking in order to work properly in a multithreaded application. It looks like this locking is only required due to the use of the c_warpMat. This will also affect the Affine transforms.
<pre>
namespace cv { namespace gpu { namespace device
{
namespace imgproc
{
__constant__ float c_warpMat[3 * 3];
struct AffineTransform
{
static __device__ __forceinline__ float2 calcCoord(int x, int y)
{
const float xcoo = c_warpMat[0] * x + c_warpMat[1] * y + c_warpMat[2];
const float ycoo = c_warpMat[3] * x + c_warpMat[4] * y + c_warpMat[5];
return make_float2(xcoo, ycoo);
}
};
</pre>
```
## History
##### Nikita Manovich on 2013-12-11 07:56
```
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: gpu/cuda (contrib) | low | Critical |
97,428,829 | opencv | GoodFeaturesToTrackDetector_GPU is ignoring mask when computing threshold | Transferred from http://code.opencv.org/issues/3425
```
|| Ronald Mallet on 2013-12-07 02:08
|| Priority: Normal
|| Affected: 2.4.7 (latest release)
|| Category: gpu (cuda)
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: Any / Any
```
## GoodFeaturesToTrackDetector_GPU is ignoring mask when computing threshold
```
In cv::gpu::GoodFeaturesToTrackDetector_GPU::operator(), once the corner's score are computed, it needs to find the max value to set the threshold.
However, the mask given as a parameter is not passed to the minMax function (it sends an empty one).
So if there's a maximum score outside the valid mask, that's much higher than anything within the mask, then no corners are returned.
Simple fix is to pass the mask to the minMax function.
```
## History
##### Nikita Manovich on 2013-12-10 09:24
```
Hi,
Thank you for the bug report. Could you please help us to fix the problem and send us a patch for the bug? Please look at the article: http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute
- Assignee set to Ronald Mallet
- Status changed from New to Open
```
##### Nikita Manovich on 2013-12-10 11:21
```
- Category set to gpu (cuda)
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: gpu/cuda (contrib) | low | Critical |
97,428,872 | opencv | libv4l2: error converting / decoding frame data: v4l-convert: error destination buffer too small | Transferred from http://code.opencv.org/issues/3427
```
|| Patrick Brown on 2013-12-09 22:09
|| Priority: Normal
|| Affected: branch 'master' (3.0-dev)
|| Category: highgui-camera
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Linux
```
## libv4l2: error converting / decoding frame data: v4l-convert: error destination buffer too small
```
FPS: 11, width: 640, height: 480
width:3264:height:2448
HIGHGUI ERROR: V4L: Property <unknown property string>(6) not supported by device
HIGHGUI ERROR: V4L: Property <unknown property string>(8) not supported by device
FPS: 11, width: 3264, height: 2448
iteration:0
libv4l2: error converting / decoding frame data: v4l-convert: error destination buffer too small (16777216 < 23970816)
VIDIOC_DQBUF: Bad address
```
## History
##### Patrick Brown on 2013-12-09 22:10
```
This is an econ 8 megapixel camera
```
##### Patrick Brown on 2013-12-10 00:20
```
I noticed that in lib_jasper that the value 16 * 1024 * 1024 corresponds to the value 16777216 .
.//3rdparty/libjasper/jasper/jas_image.h:#define JAS_IMAGE_INMEMTHRESH (16 * 1024 * 1024)
I the error related to this define?
I rebuilt the libjasper.so.1 and installed it on my system and I get the same error so I believe that the error is the result of something else.
Has anyone ever used a 8 megapixel camera with opencv?
```
##### Nikita Manovich on 2013-12-10 19:39
```
Hi,
Thank you for the bug report. Could you please help us to investigate and fix the problem? You can find all necessary information in the article: http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute. Also please provide more information: model of the camera, exact API call which leads to the problem.
- Assignee set to Patrick Brown
```
##### Patrick Brown on 2013-12-10 21:21
```
Here is the code that creates the error:
/*
* capcam.cpp
*
* Created on: Nov 4, 2013
* Author: student
*/
#include <iostream>
#include <opencv2/opencv.hpp>
using namespace std;
using namespace cv;
#include <stdio.h>
int saveImageToFile(Mat img,char *fname);
static int maxcams = 1;
int main()
{
char fName[256];
int icnt = 0;
// VideoCapture capture(-1); // use webcam
//VideoCapture capture("input.mpg"); // use input file
Mat frame;
int icam ;
icam = 0;
while(true)
{
char cOut[256];
VideoCapture *capture;
capture = new VideoCapture(icam); // use webcam
if(capture->isOpened())
{
int mode;
// if using webcam:
int fps = 11;
// if using input file:
//int fps = capture.get(CV_CAP_PROP_FPS);
int width = capture->get(CV_CAP_PROP_FRAME_WIDTH);
int height = capture->get(CV_CAP_PROP_FRAME_HEIGHT);
cout << "FPS: " << fps << ", width: " << width << ", height: " << height << endl;
width = 3264;//2592;//3296;//2592;
height = 2448;//1944;//2472;//1944;
capture->set(CV_CAP_PROP_FRAME_WIDTH,width);
capture->set(CV_CAP_PROP_FRAME_HEIGHT,height);
width = capture->get(CV_CAP_PROP_FRAME_WIDTH);
height = capture->get(CV_CAP_PROP_FRAME_HEIGHT);
printf("width:%d:height:%d\r\n",width,height);
mode = ('M' & 255) + (('J' & 255) << 8) + (('P' & 255) << 16) + (('G' & 255) << 24);
capture->set(CV_CAP_PROP_FOURCC, CV_FOURCC('M', 'J', 'P', 'G'));
capture->set(CV_CAP_PROP_FORMAT , mode);
cout << "FPS: " << fps << ", width: " << width << ", height: " << height << endl;
capture->set(CV_CAP_PROP_FPS, 1);
sprintf(cOut,"mycam:%d",icam);
for(int ii=0;ii<maxcams;ii++)
{
for(int jj=0;jj<10;jj++)
{
printf("iteration:%d\r\n",jj);
*capture >> frame;
}
printf("display image\r\n");
imshow(cOut,frame);
sprintf(fName,"f%d.jpg",icnt);
printf("at 1\r\n");
saveImageToFile(frame,fName);
printf("Frame to file %s\r\n",fName);
icnt = icnt + 1;
int key = waitKey(10);
if((char)key == 'q') { break; }
}
destroyWindow(cOut);
capture->release();
delete capture;
icam = icam + 1;
if(icam >= maxcams)
{
icam = 0;
}
}
}
return 0;
}
#define COMPRESSION_QUALITY (100) /* from 1 to 100 [100 is best] */
int saveImageToFile(Mat img,char *fname)
{
// Mat img(x, y, CV_16UC3, Scalar(0,50000, 50000)); //create an image ( 3 channels, 16 bit image depth, 650 high, 600 wide, (0, 50000, 50000) assigned for Blue, Green and Red plane respectively. )
if (img.empty()) //check whether the image is loaded or not
{
printf("ERROR : Image cannot be loaded..!!\r\n");
return -1;
}
else
{
printf("saving image>%s<\r\n",fname);
}
vector<int> compression_params; //vector that stores the compression parameters of the image
compression_params.push_back(CV_IMWRITE_JPEG_QUALITY); //specify the compression technique
compression_params.push_back(COMPRESSION_QUALITY); //specify the compression quality
bool bSuccess = imwrite(fname, img, compression_params); //write the image to file
if ( !bSuccess )
{
printf("ERROR : Failed to save the image\r\n");
}
else
{
printf("saved image>%s<\r\n",fname);
}
return 0;
}
```
##### Patrick Brown on 2013-12-10 21:25
```
The camera is a 8 Megapixel e-con camera that support 11 frames/sec at 8 megapixel. You can get the raw bmp image if you want.
```
##### Patrick Brown on 2013-12-11 04:19
```
In using an e-con 8 megapixel camera I am getting an error:
FPS: 11, width: 640, height: 480
width:3264:height:2448
HIGHGUI ERROR: V4L: Property <unknown property string>(6) not supported by device
HIGHGUI ERROR: V4L: Property <unknown property string>(8) not supported by device
FPS: 11, width: 3264, height: 2448
iteration:0
libv4l2: error converting / decoding frame data: v4l-convert: error destination buffer too small (16777216 < 23970816)
VIDIOC_DQBUF: Bad address
I believe that I have tracked it down to the define :
/lib/libv4l2/libv4l2-priv.h:#define V4L2_FRAME_BUF_SIZE (4096 * 4096)
I am trying to contact the support group for this code.
```
##### Nikita Manovich on 2013-12-11 08:10
```
- Status changed from New to Open
```
##### Patrick Brown on 2013-12-11 20:19
```
I had made a change to http://git.linuxtv.org/v4l-utils.git/tree tip in file
v4l-utils-HEAD-4dea4af/lib/libv4l2/libv4l2-priv.h
The change was
define V4L2_FRAME_BUF_SIZE (4096 * 4096)
to
#define V4L2_FRAME_BUF_SIZE (2 * 4096 * 4096)
When the package is recompiled and then the shared object files updated the package now works with the 8 megapixel camera in the raw mode.
What action do you want me to take or maybe you have more insight into the how this is to work?
```
##### Vadim Pisarevsky on 2015-05-30 07:02
```
- Affected version changed from 2.4.0 - 2.4.6 to branch 'master'
(3.0-dev)
- Category changed from highgui-gui to highgui-camera
```
| bug,auto-transferred,priority: normal,category: videoio(camera),affected: 3.4 | low | Critical |
97,428,922 | opencv | 2.4.7 for iOS Not Running on iOS 6 | Transferred from http://code.opencv.org/issues/3435
```
|| Kevin Clark on 2013-12-13 20:50
|| Priority: Normal
|| Affected: branch '2.4' (2.4-dev)
|| Category: ios
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / iOS
```
## 2.4.7 for iOS Not Running on iOS 6
```
We have been using a much older version of OpenCV in our iOS app for quite some time with no problems. This older library has worked on iOS4-7. Recently we started integrating the latest openCV (2.4.7) and have found that it no longer runs on iOS 6. We only support iOS6 and 7 now so I can’t speak for earlier version of the iOS. The project builds just fine, but once it starts actually running on the device we get this error:
<pre>
dyld: lazy symbol binding failed: Symbol not found: ___sincos_stret Referenced from: /var/mobile/Applications/E52CFE77-395C-4653-BF70-0A412A734524/MyProject.app/MyProject Expected in: /usr/lib/libSystem.B.dylib
dyld: Symbol not found: ___sincos_stret Referenced from: /var/mobile/Applications/E52CFE77-395C-4653-BF70-0A412A734524/MyProject.app/MyProject Expected in: /usr/lib/libSystem.B.dylib
</pre>
Again, this is ONLY seen on iOS 6 devices. We've tested on an iPhone 4 and iPhone 5 with the same problem. OpenCV is the only thing we have changed since we started seeing this error so it is localized to OpenCV.
```
## History
##### Alexander Shishkov on 2013-12-17 06:52
```
- Assignee set to Alexander Shishkov
- Category set to ios
```
##### Anna Kogan on 2014-01-13 09:15
```
- Affected version changed from 2.4.7 (latest release) to branch '2.4'
(2.4-dev)
- Description changed from We have been using a much older version of
OpenCV in our iOS app for quite so... to We have been using a much
older version of OpenCV in our iOS app for quite so... More
```
##### Andrew Senin on 2014-01-21 10:45
```
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,affected: 2.4,platform: ios/osx | low | Critical |
97,429,001 | opencv | Ptr_CalibrateDebevec is not defined in the generated file by matlab code generator | Transferred from http://code.opencv.org/issues/3445
```
|| Yantao Xie on 2013-12-19 06:57
|| Priority: Normal
|| Affected: branch 'master' (3.0-dev)
|| Category: matlab bindings
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Linux
```
## Ptr_CalibrateDebevec is not defined in the generated file by matlab code generator
```
I compiled the current master branch on Linux. I've got the following errors during compiling the matlab module,
<pre>
CMake Error at /home/xxx/hisstone/opencv/modules/matlab/compile.cmake:47 (message):
Failed to compile createCalibrateDebevec:
/home/xxx/hisstone/opencv/build/modules/matlab/src/createCalibrateDebevec.cpp:
In function ‘void mexFunction(int, mxArray**, int, const mxArray**)’:
/home/xxx/hisstone/opencv/build/modules/matlab/src/createCalibrateDebevec.cpp:46:3:
error: ‘Ptr_CalibrateDebevec’ was not declared in this scope
</pre>
I think there missing a statement before the mex function,
@
typedef Ptr<CalibrateDebevec> Ptr_CalibrateDebevec;
@
But this file is generated by python generator and I don't know how to update it.
```
## History
##### Alexander Shishkov on 2013-12-20 12:37
```
- Assignee set to Alexander Shishkov
- Category set to photo
```
##### Sergey Zagoruyko on 2013-12-22 14:52
```
brigde.hpp misses photo typedefs and operators.
I did not know how to edit python generator code too so I tried to add the following:
@typedef cv::Ptr<cv::CalibrateDebevec> Ptr_CalibrateDebevec;
typedef cv::Ptr<cv::CalibrateRobertson> Ptr_CalibrateRobertson;
typedef cv::Ptr<cv::LineSegmentDetector> Ptr_LineSegmentDetector;
typedef cv::Ptr<cv::MergeDebevec> Ptr_MergeDebevec;
typedef cv::Ptr<cv::MergeMertens> Ptr_MergeMertens;
typedef cv::Ptr<cv::MergeRobertson> Ptr_MergeRobertson;
typedef cv::Ptr<cv::Tonemap> Ptr_Tonemap;
typedef cv::Ptr<cv::TonemapDrago> Ptr_TonemapDrago;
typedef cv::Ptr<cv::TonemapDurand> Ptr_TonemapDurand;
typedef cv::Ptr<cv::TonemapMantiuk> Ptr_TonemapMantiuk;
typedef cv::Ptr<cv::TonemapReinhard> Ptr_TonemapReinhard;@
and
@ Bridge& operator=(const Ptr<AlignMTB>& ) { return *this; }
Bridge& operator=(const Ptr<CalibrateDebevec>& ) { return *this; }
Bridge& operator=(const Ptr<CalibrateRobertson>& ) { return *this; }
Bridge& operator=(const Ptr<LineSegmentDetector>& ) { return *this; }
Bridge& operator=(const Ptr<MergeDebevec>& ) { return *this; }
Bridge& operator=(const Ptr<MergeMertens>& ) { return *this; }
Bridge& operator=(const Ptr<MergeRobertson>& ) { return *this; }
Bridge& operator=(const Ptr<Tonemap>& ) { return *this; }
Bridge& operator=(const Ptr<TonemapDrago>& ) { return *this; }
Bridge& operator=(const Ptr<TonemapDurand>& ) { return *this; }
Bridge& operator=(const Ptr<TonemapMantiuk>& ) { return *this; }
Bridge& operator=(const Ptr<TonemapReinhard>& ) { return *this; }@
Reported on Mac OS 10.9 and Ubuntu 12.10
It worked and now I some new errors in findCirclesGrid, but it looks unrelated to this issue
```
##### Donald Venable on 2013-12-23 22:49
```
I was having the same issue, Sergey's fix got me to the same point that he's at. Im compiling on OS X 10.9 (Mavericks), XCode 5.
Here's my current Error Set:
<pre>
Failed to compile findCirclesGrid:
/Users/venabled/Downloads/opencv/build/modules/matlab/src/findCirclesGrid.cpp:46:56:
error: conditional expression is ambiguous; 'Ptr<class
cv::SimpleBlobDetector>' can be converted to 'Ptr<cv::FeatureDetector>' and
vice versa
Ptr_FeatureDetector blobDetector = inputs[3].empty() ? makePtr<SimpleBlobDetector>() : inputs[3].toPtrFeatureDetector();
^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from
/Users/venabled/Downloads/opencv/build/modules/matlab/src/findCirclesGrid.cpp:14:
In file included from
/Users/venabled/Downloads/opencv/modules/matlab/include/opencv2/matlab/bridge.hpp:49:
In file included from
/Users/venabled/Downloads/opencv/modules/core/include/opencv2/core.hpp:52:
In file included from
/Users/venabled/Downloads/opencv/modules/core/include/opencv2/core/base.hpp:54:
In file included from
/Users/venabled/Downloads/opencv/modules/core/include/opencv2/core/cvstd.hpp:906:
/Users/venabled/Downloads/opencv/modules/core/include/opencv2/core/ptr.inl.hpp:139:48:
error: cannot initialize a member subobject of type 'cv::SimpleBlobDetector
*' with an lvalue of type 'cv::FeatureDetector *const'
Ptr<T>::Ptr(const Ptr<Y>& o) : owner(o.owner), stored(o.stored)
^ ~~~~~~~~
/Users/venabled/Downloads/opencv/build/modules/matlab/src/findCirclesGrid.cpp:46:90:
note: in instantiation of function template specialization
'cv::Ptr<cv::SimpleBlobDetector>::Ptr<cv::FeatureDetector>' requested here
Ptr_FeatureDetector blobDetector = inputs[3].empty() ? makePtr<SimpleBlobDetector>() : inputs[3].toPtrFeatureDetector();
^
2 errors generated.
mex: compile of ' "/Users/venabled/Downloads/opencv/build/modules/matlab/src/findCirclesGrid.cpp"' failed.
</pre>
```
##### Sergey Zagoruyko on 2013-12-27 13:19
```
Fixed it by editing modules/matlab/src/findCirclesGrid.cpp
@// Ptr_FeatureDetector blobDetector = inputs[3].empty() ? makePtr<SimpleBlobDetector>() : inputs[3].toPtrFeatureDetector();
Ptr_FeatureDetector blobDetector;
if (inputs[3].empty())
blobDetector = makePtr<SimpleBlobDetector> ();
else
blobDetector = inputs[3].toPtrFeatureDetector();@
was able to compile but now I have a new error
@>> cv.Rodrigues(eye(3))
Error using cv.Rodrigues
Invalid MEX-file '/opt/opencv/modules/matlab/+cv/Rodrigues.mexa64':
/opt/local/MATLAB/R2013a/bin/glnxa64/../../sys/os/glnxa64/libstdc++.so.6: version `GLIBCXX_3.4.15' not found (required by
/opt/opencv/lib/libopencv_calib3d.so.3.0)@
```
##### Hilton Bristow on 2013-12-29 14:28
```
I think that type is new in 3.0, and as such I haven't added a definition for it yet (See #3465).
Sergey, that libc++ error that you're seeing is non-trivial to solve. Matlab ships its own versions of the standard library (and OpenCV), and preferences them over your system versions in almost all cases. This can cause some dynamically-loaded library hell, and I'm yet to find a bulletproof method of fixing this on all platforms.
```
##### Sergey Zagoruyko on 2013-12-31 12:44
```
Hi Hilton,
Could you please add it?
Finally I was able to run bindings on Mac OS 10.9 & Xcode 5, Ubuntu is still stuck. Anyways it looks impossible to compile bindings in UNIX without additional mexopts.sh setup and manual fixes.
There is another bug at modules/matlab/CMakeLists.txt:107
@prepend("-L" MEX_LIB_DIR ${LIBRARY_OUTPUT_PATH}/$<CONFIGURATION>)@
looks like /$<CONFIGURATION> prefix should work in Windows but not UNIX, there should be an if statement for different platforms. Sorry I cannot do pull requests.
```
##### Anna Kogan on 2014-01-13 09:31
```
- Status changed from New to Open
```
##### Riddhiman Dasgupta on 2014-02-28 10:19
```
Sergey Zagoruyko wrote:
> brigde.hpp misses photo typedefs and operators.
> I did not know how to edit python generator code too so I tried to add the following:
>
> @typedef cv::Ptr<cv::CalibrateDebevec> Ptr_CalibrateDebevec;
> typedef cv::Ptr<cv::CalibrateRobertson> Ptr_CalibrateRobertson;
> typedef cv::Ptr<cv::LineSegmentDetector> Ptr_LineSegmentDetector;
> typedef cv::Ptr<cv::MergeDebevec> Ptr_MergeDebevec;
> typedef cv::Ptr<cv::MergeMertens> Ptr_MergeMertens;
> typedef cv::Ptr<cv::MergeRobertson> Ptr_MergeRobertson;
> typedef cv::Ptr<cv::Tonemap> Ptr_Tonemap;
> typedef cv::Ptr<cv::TonemapDrago> Ptr_TonemapDrago;
> typedef cv::Ptr<cv::TonemapDurand> Ptr_TonemapDurand;
> typedef cv::Ptr<cv::TonemapMantiuk> Ptr_TonemapMantiuk;
> typedef cv::Ptr<cv::TonemapReinhard> Ptr_TonemapReinhard;@
>
> and
>
> @ Bridge& operator=(const Ptr<AlignMTB>& ) { return *this; }
> Bridge& operator=(const Ptr<CalibrateDebevec>& ) { return *this; }
> Bridge& operator=(const Ptr<CalibrateRobertson>& ) { return *this; }
> Bridge& operator=(const Ptr<LineSegmentDetector>& ) { return *this; }
> Bridge& operator=(const Ptr<MergeDebevec>& ) { return *this; }
> Bridge& operator=(const Ptr<MergeMertens>& ) { return *this; }
> Bridge& operator=(const Ptr<MergeRobertson>& ) { return *this; }
> Bridge& operator=(const Ptr<Tonemap>& ) { return *this; }
> Bridge& operator=(const Ptr<TonemapDrago>& ) { return *this; }
> Bridge& operator=(const Ptr<TonemapDurand>& ) { return *this; }
> Bridge& operator=(const Ptr<TonemapMantiuk>& ) { return *this; }
> Bridge& operator=(const Ptr<TonemapReinhard>& ) { return *this; }@
>
> Reported on Mac OS 10.9 and Ubuntu 12.10
>
> It worked and now I some new errors in findCirclesGrid, but it looks unrelated to this issue
I tried this, and it gave rise to a ton of new errors to the tune of:
@TonemapReinhard is not defined in this scope
Bridge& operator=(const int&) cannot be overloaded
error: invalid conversion from ‘cv::MergeMertens*’ to
‘cv::bridge::Ptr_MergeMertens {aka int}’ [-fpermissive]@
The findCirclesGrid file also gave a few errors, but they were resolved by using Sergey's suggestions. Compilation did not progress though, due to the above errors.
Is there something I'm doing wrong while compiling the MATLAB module?
```
##### Vadim Pisarevsky on 2015-05-14 16:54
```
- Category changed from photo to matlab bindings
```
| bug,auto-transferred,priority: normal,affected: 3.4,category: matlab bindings | low | Critical |
97,429,027 | opencv | MouseCallback gives unsigned values when mouse is outside the window | Transferred from http://code.opencv.org/issues/3446
```
|| Adi Shavit on 2013-12-19 08:06
|| Priority: Normal
|| Affected: 2.4.8 (latest release)
|| Category: highgui-gui
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Windows
```
## MouseCallback gives unsigned values when mouse is outside the window
```
When the mouse goes out of the window, e.g. while the button is still pressed, the MouseCallback event is received with the wrong coordinates.
When the mouse is above or to the left of the window, the values x and/or y (which are int) get are the unsigned short casting of the expected negative values.
Casting x/y to short returns the correct negative position.
On Windows x and y are recieved as the low and high words of a single int.
When separating them into 2 separate ints they should be cast to shorts before the assignment to allow for negative values.
See this row "here":http://code.opencv.org/projects/opencv/repository/revisions/master/entry/modules/highgui/src/window_w32.cpp#L1478 .
```
## History
##### Anna Kogan on 2014-01-13 09:19
```
Hello Adi,
Thank you for reporting the problem. If you could find on a solution for the issue, a contribution (see [[How_to_contribute]]) would be very appreciated!
- Affected version changed from 2.4.0 - 2.4.6 to 2.4.8 (latest
release)
- Assignee set to Adi Shavit
```
##### Andrew Senin on 2014-01-26 16:06
```
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,category: highgui-gui,affected: 2.4 | low | Critical |
97,429,139 | opencv | cv::ocl::OclCascadeClassifierBuf throws an exception for some combinations of flags | Transferred from http://code.opencv.org/issues/3452
```
|| Evgeniy Badaev on 2013-12-23 07:47
|| Priority: Normal
|| Affected: branch '2.4' (2.4-dev)
|| Category: ocl
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: Any / Any
```
## cv::ocl::OclCascadeClassifierBuf throws an exception for some combinations of flags
```
'detectMultiScale' method fails with an exception whenever 'flags' argument contains both 'CV_HAAR_SCALE_IMAGE' and 'CV_HAAR_FIND_BIGGEST_OBJECT'.
Exception originates from the following statement due to an empty 'gimg1':
<pre>
resizeroi = gimg1(roi2);
</pre>
The issue is caused by an inconsistent 'flags' handling in 'detectMultiScale' and 'Init' methods: 'Init' method strips 'CV_HAAR_SCALE_IMAGE' flags whenever 'CV_HAAR_FIND_BIGGEST_OBJECT' is present, whereas 'detectMultiScale' method contains no such handling for 'flags'.
IMO this could be fixed by moving the following statements away from 'Init' method into 'detectMultiScale' method (right before the call to 'Init'):
<pre>
findBiggestObject = (flags & CV_HAAR_FIND_BIGGEST_OBJECT) != 0;
if( findBiggestObject )
flags &= ~(CV_HAAR_SCALE_IMAGE | CV_HAAR_DO_CANNY_PRUNING);
</pre>
```
## History
##### Evgeniy Badaev on 2013-12-23 08:16
```
An alternative solution would be to replace 'flags' with 'm_flags' in the following statement of 'detectMultiScale' method:
<pre> if( (flags & CV_HAAR_SCALE_IMAGE) )</pre>
```
##### Evgeny Talanin on 2013-12-27 09:13
```
Thanks, Evgeniy!
Could нou submit a pull request containing your modifications to our github repo as described in http://code.opencv.org/projects/opencv/wiki/How_to_contribute? This will really help since fix will be automatically tested by our build system.
Of course you can also implement a test to prove your suggestion is correct and add it to your pull request as well.
- Assignee set to Evgeniy Badaev
```
##### Ilya Lavrenov on 2013-12-28 11:18
```
- Category set to ocl
```
##### Anna Kogan on 2014-01-13 09:23
```
- Affected version changed from 2.4.7 (latest release) to branch '2.4'
(2.4-dev)
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: ocl | low | Critical |
97,429,299 | opencv | CMake does not correctly find ilmbase. | Transferred from http://code.opencv.org/issues/3456
```
|| Robert Minsk on 2013-12-24 00:07
|| Priority: Low
|| Affected: 2.4.0 - 2.4.7
|| Category: build/install
|| Tracker: Bug
|| Difficulty: Easy
|| PR:
|| Platform: x86 / Linux
```
## CMake does not correctly find ilmbase.
```
In our facility ilmbase is installed in a separate diectory from openexr. The file opencv-2.4.7/cmake/OpenCVFindOpenEXR.cmake assumes ilmbase is in the same directory as openexr. Attached is a patch to fix the cmake file.
```
## History
##### Anna Kogan on 2014-01-13 09:09
```
Hello Robert,
Thank you for reporting the problem and especially for providing the solution for it!
OpenCV is developed in "GitHub repo":https://github.com/Itseez/opencv, and all the contributions are merged in the library only by making a pull request. Would you mind making one as described on the wiki page: [[How_to_contribute]]?
- Affected version changed from 2.4.7 (latest release) to 2.4.0 -
2.4.7
```
##### Andrew Senin on 2014-01-23 14:15
```
- Assignee changed from Roman Donchenko to Robert Minsk
- Status changed from New to Open
```
| bug,auto-transferred,priority: low,category: build/install,affected: 2.4 | low | Critical |
97,429,412 | opencv | Type errors in Matlab module's bridge.hpp when compiling without some optional modules (calib3d, etc.) | Transferred from http://code.opencv.org/issues/3464
```
|| Hilton Bristow on 2013-12-29 14:17
|| Priority: Normal
|| Affected: branch 'master' (3.0-dev)
|| Category: matlab bindings
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: Any / Any
```
## Type errors in Matlab module's bridge.hpp when compiling without some optional modules (calib3d, etc.)
```
This issue is for my own records so I can put an issue # next to my pull request when I fix it.
---------------------------------------
When compiling OpenCV without some optional modules (calib3d, etc.) the Matlab bridge fails to find certain includes required for type conversions. Either guard these includes and their corresponding types/conversions, or hard-code paths to all of the top-level module includes.
<pre>
Failed to compile absdiff: In file included from
/Users/hiltonbristow/Code/projects/opencv/build/modules/matlab/src/absdiff.cpp:14:
/Users/hiltonbristow/Code/projects/opencv/modules/matlab/include/opencv2/matlab/bridge.hpp:51:10:
fatal error: 'opencv2/calib3d.hpp' file not found
#include <opencv2/calib3d.hpp>
^
1 error generated.
</pre>
```
## History
##### Anna Kogan on 2014-01-13 09:24
```
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,affected: 3.4,category: matlab bindings | low | Critical |
97,429,468 | opencv | Many type-conversions missing in the Matlab module | Transferred from http://code.opencv.org/issues/3465
```
|| Hilton Bristow on 2013-12-29 14:20
|| Priority: Normal
|| Affected: branch 'master' (3.0-dev)
|| Category: matlab bindings
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: Any / Any
```
## Many type-conversions missing in the Matlab module
```
This issue is for my own records so I can put an issue # next to my pull request when I fix it.
---------------------------------------
Many type conversions in the Matlab module are still just stubs, and cause runtime exceptions in OpenCV functions that require them. Implement these, and add newly-introduced types for 3.0
```
## History
##### Anna Kogan on 2014-01-13 09:24
```
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,affected: 3.4,category: matlab bindings | low | Critical |
97,429,598 | opencv | The orientation of camera preview is different on IOS6 with IOS7 | Transferred from http://code.opencv.org/issues/3473
```
|| Gang Tong on 2014-01-03 02:57
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.7
|| Category: ios
|| Tracker: Bug
|| Difficulty: Easy
|| PR:
|| Platform: ARM / iOS
```
## The orientation of camera preview is different on IOS6 with IOS7
```
The default orientation of camera preview on IOS6 is portrait, but is landscape on IOS7.
```
## History
##### Gang Tong on 2014-01-07 04:38
```
- Assignee set to Vadim Pisarevsky
```
##### Alexander Smorkalov on 2014-01-19 12:13
```
- Target version set to 2.4.9
- Category set to ios
```
##### Andrew Senin on 2014-01-20 14:29
```
Hello Gang,
Unfortunately our resources are limited at the moment. Do you think you could work on a solution on your own? Your contribution to the OpenCV repo (see http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute) would be very appreciated!
- Difficulty changed from Medium to Easy
- Assignee changed from Vadim Pisarevsky to Gang Tong
- Status changed from New to Open
```
##### Alexander Smorkalov on 2014-04-30 19:05
```
- Target version changed from 2.4.9 to 2.4.10
```
| bug,auto-transferred,priority: normal,affected: 2.4,platform: ios/osx | low | Critical |
97,429,648 | opencv | Add ability to serve video stream | Transferred from http://code.opencv.org/issues/3475
```
|| Cody Rigney on 2014-01-06 21:56
|| Priority: Normal
|| Affected: None
|| Category: highgui-video
|| Tracker: Feature
|| Difficulty: Hard
|| PR:
|| Platform: None / None
```
## Add ability to serve video stream
```
Could the capability be added to easily broadcast a motion jpeg to clients using VideoWriter? Does it make sense to add this to video writer, or it's own class? If added to VideoWriter, additional parameters could be passed for binding IP address and port number.
This feature would make debugging on embedded devices, for example, much easier.
It seems to make sense for Issue #1324 to be completed before this one. I can complete these, but I need to find a couple answers to some questions first.
```
## History
##### Cody Rigney on 2014-01-06 21:58
```
Is there a class used for networking (across platforms) in the opencv source? Or at least an example? I will dig through and look if I don't get an answer. Thanks.
```
##### Andrew Senin on 2014-01-23 14:13
```
Here is a link I found:
http://www.rosoo.net/a/201004/9123.html
Haven't tried, though.
- Difficulty set to Hard
- Target version changed from 3.0 to Next Hackathon
- Assignee deleted (Cody Rigney)
- Status changed from New to Open
```
##### Cody Rigney on 2014-01-24 02:01
```
Thanks.
I have this implemented for windows and unix-like systems. It's single threaded but can serve multiple clients. It only does mjpeg right now which means you can view it from a browser. (Great for embedded device debugging). All you do is open a videowriter but specify a port and the fourcc for mjpeg only. Then write frames as usual. I still have to test the unix part and builds then I will submit a pull request.
```
##### Andrew Senin on 2014-01-24 10:02
```
Thanks Cody! That would be great!
- Target version changed from Next Hackathon to 3.0-alpha
- Assignee set to Cody Rigney
```
##### be rak on 2014-05-30 09:49
```
any updates on this ? a lot of people have been asking
```
##### Cody Rigney on 2014-08-04 14:24
```
Sorry for the long delay. I had the code done a while back but bought a new computer and forgot about that enhancement I had been working on, on the previous computer. I will try to get this committed within the next couple of weeks. If not, I'll post what I have so far for another person to finish.
```
##### Cody Rigney on 2014-08-04 14:28
```
Another point to this, I have noticed that the popular browsers are bugged on what I implemented (multi-part jpeg). Firefox has a massive memory leak but works, while Chrome and IE require it to be embedded in HTML to work. Either way, I'll try to finish it up.
```
##### Aly Osama on 2014-12-17 19:44
```
Hi, i want to ask what is the progress in this feature so that i could complete it
thanks
```
##### Cody Rigney on 2015-03-02 15:02
```
Aly Osama wrote:
> Hi, i want to ask what is the progress in this feature so that i could complete it
> thanks
Hi, I'll submit the patch. I thought I would have the chance to complete it, but I have been very busy. Please give me till the end of the week. Thanks.
```
##### Maksim Shabunin on 2015-04-27 09:11
```
- Target version changed from 3.0-alpha to 3.0
```
| auto-transferred,priority: normal,feature,category: videoio | low | Critical |
97,429,754 | opencv | highgui.CV_CAP_PROP_FOURCC returns 0 with certain VideoFiles | Transferred from http://code.opencv.org/issues/3487
```
|| Peter Meier on 2014-01-15 16:14
|| Priority: Normal
|| Affected: branch 'master' (3.0-dev)
|| Category: java bindings
|| Tracker: Bug
|| Difficulty: Easy
|| PR:
|| Platform: x64 / Windows
```
## highgui.CV_CAP_PROP_FOURCC returns 0 with certain VideoFiles
```
Hi,
I'm using OpenCV-java(3.00) (selfcompiled version see issue #3460)
and the CV_CAP_PROP_FOURCC property gives me back 0 with certain Video formats.
I've tested Mpg-2,flash(.flv) and mp4 until now.
It opens every one of them, but only the .mp4 gave another value than 0-(avc1).
So I do not no if it is normal, or (OpenCV-)java specific..
```
## History
##### Andrew Senin on 2014-01-24 08:59
```
Peter,
Thanks for posting the issue. As you mentioned in the #3460 not all HighGUI functionality is covered by the Java wrappers.
Unfortunately currently there are no people supporting the Java wrappers on regular basis. So you can help the project by contributing your fix. Please read more details on how to contribute here: http://code.opencv.org/projects/opencv/wiki/How_to_contribute
Thanks
- Target version changed from 3.0-beta to Next Hackathon
- Difficulty changed from Medium to Easy
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,affected: 3.4,category: java bindings | low | Critical |
97,429,784 | opencv | SIGSEGV before entering main() when module viz is compiled into OpenCV | Transferred from http://code.opencv.org/issues/3492
```
|| David W. on 2014-01-20 16:59
|| Priority: Normal
|| Affected: branch 'master' (3.0-dev)
|| Category: viz
|| Tracker: Bug
|| Difficulty: Medium
|| PR:
|| Platform: x64 / Linux
```
## SIGSEGV before entering main() when module viz is compiled into OpenCV
```
I wanted to try out the new viz module, so i run cmake with '-D WITH_VTK=ON'.
The building process is completed properly, but if i link a program against OpenCV, it crashes with SIGSEGV before entering main().
My debugger stops somewhere in /usr/lib/libQtGui.so.4, like here: http://www.qtcentre.org/threads/54158-OpenCV-Seg-fault
Stacktrace:
@Thread 1 (Thread 0x7ffff7f827c0 (LWP 27951)):
#0 0x00007fffe550631c in ?? () from /usr/lib/libQtGui.so.4
No symbol table info available.
#1 0x00007ffff7deb8da in call_init.part.0 () from /lib64/ld-linux-x86-64.so.2
No symbol table info available.
#2 0x00007ffff7deb9c3 in _dl_init_internal () from /lib64/ld-linux-x86-64.so.2
No symbol table info available.
#3 0x00007ffff7dde29a in _dl_start_user () from /lib64/ld-linux-x86-64.so.2
No symbol table info available.
#4 0x0000000000000002 in ?? ()
No symbol table info available.
#5 0x00007fffffffea42 in ?? ()
No symbol table info available.
#6 0x00007fffffffea6d in ?? ()
No symbol table info available.
#7 0x0000000000000000 in ?? ()
No symbol table info available.
@
I have Qt5 aswell as Qt4 installed on my machine. Cmake outputs:
@-- GUI:
-- QT 5.x: YES (ver 5.2.0)
-- QT OpenGL support: NO
-- OpenGL support: NO
-- VTK support: YES (ver 5.10.1)
@
Without viz there are no problems. I hope this is enough information.
Looks like viz is linked against the wrong version of Qt?
```
## History
##### David W. on 2014-01-20 17:06
```
- Assignee set to Anatoly Baksheev
```
##### David W. on 2014-01-20 19:16
```
I added '-D WITH_OPENGL=ON' to cmake, so now i get:
@-- GUI:
-- QT 5.x: YES (ver 5.2.0)
-- QT OpenGL support: YES (Qt5::OpenGL 5.2.0)
-- OpenGL support: YES (/lib64/libGLU.so /lib64/libGL.so /lib64/libSM.so /lib64/libICE.so /lib64/libX11.so /lib64/libXext.so)
-- VTK support: YES (ver 5.10.1)
@
But the problem remains.
```
##### Anatoly Baksheev on 2014-01-20 21:00
```
Thank you for the reporting. The OpenGL flag has no effect there, it is related to another functionality.
I don't have understanding why it happerns, need to investigate. Any help appreciated.
Now I can only suggest you trying to donwload VTK 6.0 sources and compile it with Qt5. For this you need to switch coresponding Qt5 flag in VTK cmake, otherwise it will continue to search qt4.
- Target version set to 3.0-alpha
- Difficulty changed from Easy to Medium
- Category set to viz
```
##### Anatoly Baksheev on 2014-01-20 21:01
```
Does the crash happen even for console applications that don't use qt?
```
##### David W. on 2014-01-20 23:39
```
Anatoly Baksheev wrote:
> Does the crash happen even for console applications that don't use qt?
Yes. I commented out every use of highgui or viz in my project, which is build with cmake.
- Target version deleted (3.0-alpha)
```
##### David W. on 2014-01-20 23:49
```
Anatoly Baksheev wrote:
> Thank you for the reporting. The OpenGL flag has no effect there, it is related to another functionality.
>
> I don't have understanding why it happerns, need to investigate. Any help appreciated.
>
> Now I can only suggest you trying to donwload VTK 6.0 sources and compile it with Qt5. For this you need to switch coresponding Qt5 flag in VTK cmake, otherwise it will continue to search qt4.
I had a quick look on the CMakeLists of VTK and can't find any central QT5 switch. @grep -r QT5 VTK6.0.0@ gives no results.
If you can point me to, i will try that tomorrow.
```
##### Anatoly Baksheev on 2014-01-21 08:28
```
Here is howto:
https://coderwall.com/p/akk-og
Probably only solution for this is to prevent adding opnecv_viz.so to ${OpenCV_LIBS} in case OpenCV is compiled with Qt5.
```
##### Andrew Senin on 2014-01-21 10:14
```
- Status changed from New to Open
```
##### David W. on 2014-01-22 13:03
```
Anatoly Baksheev wrote:
> Here is howto:
> https://coderwall.com/p/akk-og
>
> Probably only solution for this is to prevent adding opnecv_viz.so to ${OpenCV_LIBS} in case OpenCV is compiled with Qt5.
The original problem was solved by building viz against VTK 6.1, which required a change in OpenCVDetectVTK.cmake. My project compiles and runs properly, until I try to create a viz-window:
@ERROR: In vtk-git/src/VTK/Rendering/OpenGL/vtkOpenGLActor.cxx, line 94
vtkOpenGLActor (0x14679a0): failed after Render 1 OpenGL errors detected
0 : (1284) Stack underflow@
```
##### Anatoly Baksheev on 2014-01-23 11:08
```
Thank very much. I know this occasional issue. I think this requires investigation and some work to add support of newer VTKs. It's TODO for me when I have time (I spend my personal time on it)
Now I will only push guard that prevents linking Viz if OpenCV is compiled with qt5.
```
##### Anatoly Baksheev on 2014-01-26 16:52
```
I can't reproduce the OpenGL error on my Ubuntu 12.04 64bit. I compiled both OpenCV and VTK from git (6.2 and 6.1) with/without Qt5.2. Everything just works for me. I tried debug and release configurations. BTW, if you run 'opencv_test_viz' executable for testing, you need to specify export OPENCV_TEST_DATA_PATH=<opencv_extra_repository_path/testdaa>.
In VTK 6.1 an OpenGL error reporting layer was added, it catches many errors which were just ignored before. That's why I have a lot of OpenGL errors on my OSX 10.9 with VTK6.2. I think VTK support of 10.9 system is immature, so plan to wait stable release before looking into this.
Regarding VTK 5.x, I learned that 'apt-get install'ed VTK depends on qt4. So, you need to recompile VTK5.10 from source without Qt4 to use it with OpenCV+qt5.
```
##### David W. on 2014-02-01 19:56
```
I am working on a current arch linux x86_64 and did a clean rebuild of my project against OpenCV/master against VTK 6.1 against Qt 5.2 and finally got a working Viz window!
I think they patched something in VTK for the final 6.1 release, that made the OpenGL error disappear. I am looking forward to try out all the nice features!
Thanks for your investigations!
```
##### Vladislav Vinogradov on 2014-02-02 12:42
```
Hello David,
Since the issue seems to be resolved, I am closing the ticket.
Please reopen it, if the crash still appears.
- Status changed from Open to Cancelled
```
##### David W. on 2014-02-03 16:26
```
David W. wrote:
> I am working on a current arch linux x86_64 and did a clean rebuild of my project against OpenCV/master against VTK 6.1 against Qt 5.2 and finally got a working Viz window!
Cheered to soon!
The crash happens if setCamera() is called. I called it like this:
@
viz::Camera vizCam(Vec2d(100,100), myWindow.getWindowSize());
myWindow.setCamera(vizCam);
myWindow.spin();
@
*Console output:*
<pre>
ERROR: In VTK/Rendering/OpenGL/vtkOpenGLActor.cxx, line 94
vtkOpenGLActor (0xc828f0): failed after Render 1 OpenGL errors detected
0 : (1284) Stack underflow
</pre>
*Stack trace of main thread:*
<pre>
0 vtkRenderer::UpdateGeometry() /lib64/libvtkRenderingCore-6.1.so.1 0x7fffd9ff257a
1 vtkOpenGLRenderer::DeviceRender() /lib64/libvtkRenderingOpenGL-6.1.so.1 0x7fffd972a44b
2 vtkRenderer::Render() /lib64/libvtkRenderingCore-6.1.so.1 0x7fffd9ff4884
3 cv::viz::Viz3d::VizImpl::setCamera(cv::viz::Camera const&) /home/davidek/ba/opencv/src/opencv/build/lib/libopencv_viz.so.3.0 0x7ffff7baf547
4 main main.cpp 91 0x41da0b
</pre>
Additionally I attach the report of the viz tests. But setCamera is no testet there.
- File test_report.txt added
```
##### David W. on 2014-02-03 16:27
```
- Status changed from Cancelled to Open
```
| bug,auto-transferred,priority: normal,affected: 3.4,category: viz | low | Critical |
97,429,822 | opencv | Streaming over RTSP | Transferred from http://code.opencv.org/issues/3494
```
|| Matheus Castro Rocha on 2014-01-21 22:49
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.7
|| Category: highgui-camera
|| Tracker: Bug
|| Difficulty: Easy
|| PR:
|| Platform: x64 / Windows
```
## Streaming over RTSP
```
Hello guys !
I have been trying to access an IP Camera through RTSP in Visual Studio 2012. Here is my code:
VideoCapture cap("rtsp://IP_ADRRESS/play1.sdp/?tcp");
if (!cap.isOpened())
{
cout << "Cannot open the video file" << endl;
return -1;
}
double dWidth = cap.get(CV_CAP_PROP_FRAME_WIDTH); double dHeight = cap.get(CV_CAP_PROP_FRAME_HEIGHT); cout << "Frame size : " << dWidth << " x " << dHeight << endl;
namedWindow("MyVideo",CV_WINDOW_AUTOSIZE);
while (1)
{
Mat frame;
bool bSuccess = cap.read(frame);
if (!bSuccess)
{
cout << "Cannot read a frame from video file" << endl;
break;
}
imshow("MyVideo", frame);
if (waitKey(30) == 27)
{
cout << "esc key is pressed by user" << endl;
break;
}
}
When I run it in debug mode, it works just fine, but when I run it in realese , it doesn't work. Any ideas ?
I was able to access the camera using only ffmpeg.
if i didn't post any useful informantion, let me know !
Thanks a lot !
```
## History
##### Andrew Senin on 2014-01-23 13:52
```
Matheus, thank you for posting this issue!
Could you please let us know what version of OpenCV you use? If it is not the latest - please try with the latest. Also, please describe how it does not work (crash or no image etc.).
Since our resources now are limited it would be greatly appreciated if you could try to debug this issue locally and send as a fixing push request. You can find all details here: http://code.opencv.org/projects/opencv/wiki/How_to_contribute
Thank you
- Difficulty changed from Hard to Easy
- Target version set to Next Hackathon
- Priority changed from High to Normal
- Assignee set to Matheus Castro Rocha
- Status changed from New to Incomplete
- Category set to highgui-camera
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: videoio(camera) | low | Critical |
97,429,855 | opencv | iOS install instructions fail if there are + or - characters in the build path | Transferred from http://code.opencv.org/issues/3495
```
|| David Dancy on 2014-01-22 04:23
|| Priority: Normal
|| Affected: 2.4.8 (latest release)
|| Category: build/install
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Mac OSX
```
## iOS install instructions fail if there are + or - characters in the build path
```
My original build path was ~/Documents/Projects/C++/OpenCV-2.4.8/opencv/build
ccmake failed with regex errors when invoked from this path.
The python build script also failed.
Initially I just renamed the path to ~/Documents/Projects/C++/OpenCV020408/opencv/build but this still failed with regex and build errors.
I had to rename the path to ~/Documents/Projects/Cplusplus/OpenCV020408/opencv/build before either tool would work properly.
HTH
```
## History
##### Roman Donchenko on 2014-01-22 08:01
```
Details, please. What errors?
```
##### Andrew Senin on 2014-01-23 13:43
```
David, please provide extended description of the errors
- Assignee changed from Roman Donchenko to David Dancy
- Status changed from New to Incomplete
```
| bug,auto-transferred,priority: normal,category: build/install,affected: 2.4 | low | Critical |
97,429,994 | opencv | VideoCapture capture.read( image ) is not blocking off main thread on Android | Transferred from http://code.opencv.org/issues/3503
```
|| Arvind Pereira on 2014-01-24 22:46
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.7
|| Category: highgui-camera
|| Tracker: Bug
|| Difficulty: Medium
|| PR:
|| Platform: ARM / Android
```
## VideoCapture capture.read( image ) is not blocking off main thread on Android
```
I am running OpenCV4Android natively to perform image processing on an Android device running Android OS v 4.2.2. This is a Tegra device. I am reading images in a separate frame grabber thread, which loops around grabbing images and then processing them. I have noticed that when the device is grabbing images, both my process as well as mediaserver consume approximately 20% CPU each. This appears to me to be an issue where OpenCV's video capture device is busy-waiting for the calls to Android's camera driver to return instead of blocking on the I/O call. This in turn is causing the process to waste massive amounts of CPU simply grabbing frames. I confirmed this by commenting out all processing (I was building image pyramids and performing background MOG processing), and found that they were contributing to only an additional 5% CPU over and above grabbing frames.
Ideally, what I would expect is that OpenCV's video capture should simply block on each frame read, thus imposing negligible load on the system beyond what it has already been consuming. i.e. on my device we should have been seeing 20% CPU load on the media-server with single-digit CPU load due to copying those frames and decoding them in videocapture.
It appears that this is an issue on OS-X when the capture device is started off in a non main thread. I suspect this is a similar bug when compiled for Android. Please fix asap - it is stealing too much CPU waiting for frames to become available!
```
## History
##### Alexander Smorkalov on 2014-01-25 18:18
```
- Priority changed from Blocker to Normal
- Assignee set to Alexander Smorkalov
```
##### Andrew Senin on 2014-01-26 15:59
```
- Status changed from New to Open
```
##### Alexander Smorkalov on 2014-04-30 19:05
```
- Target version changed from 2.4.9 to 2.4.10
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: videoio(camera) | low | Critical |
97,430,069 | opencv | Need another calcOpticalFlowPyrLK Java signature taking input List<Mat> pyramid | Transferred from http://code.opencv.org/issues/3508
```
|| Xin Guo on 2014-01-28 19:18
|| Priority: Normal
|| Affected: None
|| Category: java bindings
|| Tracker: Feature
|| Difficulty: Medium
|| PR:
|| Platform: None / None
```
## Need another calcOpticalFlowPyrLK Java signature taking input List<Mat> pyramid
```
The function of @buildOpticalFlowPyramid@ and @calcOpticalFlowPyrLK@ in OpenCV-2.4.8-android-sdk is:
<pre>
public static int buildOpticalFlowPyramid(Mat img, List<Mat> pyramid, Size winSize, int maxLevel)
public static void calcOpticalFlowPyrLK(Mat prevImg, Mat nextImg, MatOfPoint2f prevPts, MatOfPoint2f nextPts, MatOfByte status, MatOfFloat err, Size winSize, int maxLevel, TermCriteria criteria, int flags, double minEigThreshold)
</pre>
If I just use the image as the input of @calcOpticalFlowPyrLK@, not using pyramid, it works well. But If I use @buildOpticalFlowPyramid@ to build a pyramid as input of @calcOpticalFlowPyrLK@. The @List<Mat> pyramid@ is not applicable for @Mat prevImg@.
If I declare the pyramid as @vector<Mat>@, @buildOpticalFlowPyramid@ works well, but the app will be forced to stop whenever call @calcOpticalFlowPyrLK@.
Here is parts of the code.
<pre>
MatOfByte status = new MatOfByte();
int maxlevel=2;
Mat prev = new Mat(mRgba.size(),CvType.CV_8UC1);
Mat curr = new Mat(mRgba.size(),CvType.CV_8UC1);
List<Mat> pyramid1= new ArrayList<Mat>(maxlevel);
List<Mat> pyramid2= new ArrayList<Mat>(maxlevel);
Mat pyramid3 = new Mat();
Mat pyramid4 = new Mat();
MatOfPoint2f featurestracked = new MatOfPoint2f();
MatOfFloat err = new MatOfFloat();
int i,count;
TermCriteria optical_flow_termination_criteria =new TermCriteria();
optical_flow_termination_criteria.epsilon =.03;
optical_flow_termination_criteria.maxCount = 20;
Size winSize = new Size(21,21);
Imgproc.cvtColor(hp.TrackingFrame, prev, Imgproc.COLOR_RGBA2GRAY,0);
Imgproc.cvtColor(mRgba, curr, Imgproc.COLOR_RGBA2GRAY,0);
//Build pyramid
int level1= Video.buildOpticalFlowPyramid(prev, pyramid1, winSize, maxlevel);
int level2 = Video.buildOpticalFlowPyramid(curr, pyramid2, winSize, maxlevel);
pyramid3 = Converters.vector_Mat_to_Mat(pyramid1);
pyramid4 = Converters.vector_Mat_to_Mat(pyramid2);
MatOfPoint2f features_next2f = new MatOfPoint2f(hp.features_next.toArray());
Video.calcOpticalFlowPyrLK(pyramid3, pyramid4,features_next2f,featurestracked,status,err, winSize,maxlevel,optical_flow_termination_criteria, 0, 0.001);
</pre>
```
## History
##### Vladislav Vinogradov on 2014-01-29 07:47
```
- Description changed from The function of buildOpticalFlowPyramid and
calcOpticalFlowPyrLK in OpenCV-2.... to The function of
buildOpticalFlowPyramid and calcOpticalFlowPyrLK in OpenCV-2....
More
```
##### Vladislav Vinogradov on 2014-01-29 07:49
```
- Description changed from The function of buildOpticalFlowPyramid and
calcOpticalFlowPyrLK in OpenCV-2.... to The function of
@buildOpticalFlowPyramid@ and @calcOpticalFlowPyrLK@ in OpenC...
More
```
##### Vladislav Vinogradov on 2014-01-29 07:52
```
- Status changed from New to Open
- Category changed from android to java bindings
```
##### Alexander Smorkalov on 2014-01-29 12:45
```
@Andrey Pavlenko, can you comment the issue?
```
##### Xin Guo on 2014-01-29 17:55
```
FYI：when I declare the pyramid as vector<Mat>, I use the Converters.vector_Mat_to_Mat(Pyramid) as the input of calcOpticalFlowPyrLK.
Thank you very much.
```
##### Xin Guo on 2014-02-02 20:43
```
Could you please comment this issue? Or give me some info about how to do about it? Thank you.
Alexander Smorkalov wrote:
> @Andrey Pavlenko, can you comment the issue?
```
##### Andrey Pavlenko on 2014-02-04 15:56
```
Java version of this function doesn't support pyramid as input.
This is actually a feature request on adding one more @calcOpticalFlowPyrLK@ signature taking @List<Mat>@ (we don't have @InputArray@ in Java).
```
##### Andrey Pavlenko on 2014-02-04 15:59
```
- Subject changed from can not input a List<Mat> pyramid into
calcOpticalFlowPyrLK (OpenCV-2.4.8-android-sdk) to Need another
calcOpticalFlowPyrLK Java signature taking input List<Mat> pyramid
- Difficulty set to Medium
- Tracker changed from Bug to Feature
```
##### Xin Guo on 2014-02-11 15:43
```
Thank you.
Andrey Pavlenko wrote:
> Java version of this function doesn't support pyramid as input.
> This is actually a feature request on adding one more @calcOpticalFlowPyrLK@ signature taking @List<Mat>@ (we don't have @InputArray@ in Java).
```
##### Alexander Smorkalov on 2014-03-28 22:20
```
- Target version set to 3.0-alpha
```
##### Maksim Shabunin on 2015-04-27 09:11
```
- Target version changed from 3.0-alpha to 3.0
```
| auto-transferred,priority: normal,feature,category: java bindings | low | Critical |
97,430,152 | opencv | CommandLineParser and negative values | Transferred from http://code.opencv.org/issues/3512
```
|| Philippe FOUBERT on 2014-01-29 13:47
|| Priority: Normal
|| Affected: branch 'master' (3.0-dev)
|| Category: core
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: Any / Any
```
## CommandLineParser and negative values
```
The parsing done in the function @CommandLineParser::CommandLineParser(int argc, const char* const argv[], const String& keys)@ seems buggy:
* @appli.exe -j=1.2@ is treated correctly
* @appli.exe -1.2@ is not treated correctly since it removes the argument because there is not possible argument named "1.2": the "-" character is considered to be only a way to specify an argument name.
```
## History
##### Vladislav Vinogradov on 2014-01-30 07:53
```
Hello Philippe,
Thank you for reporting the issue!
Unfortunately our resources are limited at the moment.
Do you think you could work on a solution on your own?
Your contribution to the OpenCV repo (see http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute for help) would be very appreciated!
- Operating System changed from Windows to Any
- HW Platform changed from x64 to Any
- Assignee changed from Vadim Pisarevsky to Philippe FOUBERT
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,category: core,affected: 3.4 | low | Critical |
97,430,256 | opencv | Remap does not support images longer than 32k pixel | Transferred from http://code.opencv.org/issues/3522
```
|| Mustafa Teke on 2014-02-05 12:08
|| Priority: Normal
|| Affected: None
|| Category: imgproc, video
|| Tracker: Feature
|| Difficulty: Medium
|| PR:
|| Platform: None / None
```
## Remap does not support images longer than 32k pixel
```
Due to use of short data type with remap, images larger than 32k are not remapped with Remap function.
```
## History
##### Dmitry Retinskiy on 2014-02-06 12:06
```
Hi Mustafa,
could you add more details to the description?
Which exactly use of short data type you are referring to?
```
##### Dmitry Retinskiy on 2014-02-07 06:56
```
- Category set to imgproc, video
```
##### Kirill Kornyakov on 2014-02-17 08:00
```
- Status changed from New to Open
- Tracker changed from Bug to Feature
```
##### Mustafa Teke on 2014-05-01 11:28
```
I deal with large satellite images often in size of a few gigabytes.
When I needed to load a long image strip I could not remap images which are vert long (4096 x 81920)
Dmitry Retinskiy wrote:
> Hi Mustafa,
>
> could you add more details to the description?
> Which exactly use of short data type you are referring to?
```
##### Andrew Senin on 2014-05-06 14:17
```
Hi Mustafa,
It would be nice to investigate this issue and make the function work with very large images. Please consider investigating this and creating a fixing pull request (http://code.opencv.org/projects/opencv/wiki/How_to_contribute). Your help is highly appreciated!
- Target version set to Next Hackathon
- Difficulty set to Medium
- Assignee set to Mustafa Teke
```
| auto-transferred,priority: normal,feature,category: imgproc,category: video | low | Critical |
97,430,286 | opencv | FileStorage Append + Memory | Transferred from http://code.opencv.org/issues/3526
```
|| Paul Triantafyllou on 2014-02-07 16:07
|| Priority: Normal
|| Affected: None
|| Category: core
|| Tracker: Feature
|| Difficulty:
|| PR:
|| Platform: None / None
```
## FileStorage Append + Memory
```
I would like to see FileStorage be able to append data to a memory buffer.
```
## History
##### Kirill Kornyakov on 2014-02-10 11:15
```
Could you please describe a scenarion when you need such functionality? May be you should implement it and submit as a pull-request?
- Assignee set to Paul Triantafyllou
- Status changed from New to Incomplete
- Category set to core
```
| auto-transferred,priority: normal,feature,category: core | low | Minor |
97,430,354 | opencv | warpPerspective causing a bus error | Transferred from http://code.opencv.org/issues/3528
```
|| Matthieu Tourne on 2014-02-07 19:53
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.7
|| Category: imgproc, video
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Mac OSX
```
## warpPerspective causing a bus error
```
Hi,
I'm loading a very large image (~600mb), then I use detector+descriptors to find a homography between another image.
Finally, I call `warpPerspective()` to apply my homography on the original image.
This causes a bus error and the error that I pasted below.
Also, calling `warpPerspective()` on the same image, but converted to a grayscale image beforehand works just fine.
I'm using opencv 2.4.7.1 on mac os 10.9, compiled with `--c++11`
Let me know if you need more information on my side!
Best,
Matthieu
<pre>
---- Error and backtrace -----
Process 17741 stopped
* thread #2: tid = 0x743d3, 0x000000010071e191 libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 263, queue = 'com.apple.root.default-priority, stop reason = EXC_BAD_ACCESS (code=2, address=0x129ab1ecc)
frame #0: 0x000000010071e191 libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 263
libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 263:
-> 0x10071e191: vmovd 3(%rcx,%rdi), %xmm3
0x10071e197: movzwl 2(%r12,%rbx,2), %edx
0x10071e19d: vmovd (%rcx,%rdi), %xmm4
0x10071e1a2: vpunpcklbw %xmm3, %xmm4, %xmm3
thread #5: tid = 0x743d6, 0x000000010071e191 libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 263, queue = 'com.apple.root.default-priority, stop reason = EXC_BAD_ACCESS (code=2, address=0x129ab1dd6)
frame #0: 0x000000010071e191 libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 263
libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 263:
-> 0x10071e191: vmovd 3(%rcx,%rdi), %xmm3
0x10071e197: movzwl 2(%r12,%rbx,2), %edx
0x10071e19d: vmovd (%rcx,%rdi), %xmm4
0x10071e1a2: vpunpcklbw %xmm3, %xmm4, %xmm3
thread #7: tid = 0x743d8, 0x000000010071e191 libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 263, queue = 'com.apple.root.default-priority, stop reason = EXC_BAD_ACCESS (code=2, address=0x129ab1ffb)
frame #0: 0x000000010071e191 libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 263
libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 263:
-> 0x10071e191: vmovd 3(%rcx,%rdi), %xmm3
0x10071e197: movzwl 2(%r12,%rbx,2), %edx
0x10071e19d: vmovd (%rcx,%rdi), %xmm4
0x10071e1a2: vpunpcklbw %xmm3, %xmm4, %xmm3
thread #8: tid = 0x743d9, 0x000000010071e26c libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 482, queue = 'com.apple.root.default-priority, stop reason = EXC_BAD_ACCESS (code=2, address=0x129ab1fb6)
frame #0: 0x000000010071e26c libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 482
libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 482:
-> 0x10071e26c: vmovd 3(%rcx,%rax), %xmm3
0x10071e272: vmovd (%rcx,%rax), %xmm4
0x10071e277: vpunpcklbw %xmm3, %xmm4, %xmm3
0x10071e27b: vpunpcklbw %xmm1, %xmm3, %xmm3
(lldb) bt
* thread #2: tid = 0x743d3, 0x000000010071e191 libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 263, queue = 'com.apple.root.default-priority, stop reason = EXC_BAD_ACCESS (code=2, address=0x129ab1ecc)
frame #0: 0x000000010071e191 libopencv_imgproc.2.4.7.dylib`cv::RemapVec_8u::operator()(cv::Mat const&, void*, short const*, unsigned short const*, void const*, int) const + 263
frame #1: 0x00000001007089f0 libopencv_imgproc.2.4.7.dylib`void cv::remapBilinear<cv::FixedPtCast<int, unsigned char, 15>, cv::RemapVec_8u, short>(cv::Mat const&, cv::Mat&, cv::Mat const&, cv::Mat const&, void const*, int, cv::Scalar_<double> const&) + 1828
frame #2: 0x000000010073254e libopencv_imgproc.2.4.7.dylib`cv::RemapInvoker::operator()(cv::Range const&) const + 2628
frame #3: 0x0000000100c9b4d4 libopencv_core.2.4.7.dylib`(anonymous namespace)::block_function(void*, unsigned long) + 120
frame #4: 0x00007fff8fc308f8 libdispatch.dylib`_dispatch_client_callout2 + 8
frame #5: 0x00007fff8fc30cca libdispatch.dylib`_dispatch_apply_serial + 42
frame #6: 0x00007fff8fc292ad libdispatch.dylib`_dispatch_client_callout + 8
frame #7: 0x00007fff8fc303cf libdispatch.dylib`_dispatch_sync_f_invoke + 39
frame #8: 0x00007fff8fc3062a libdispatch.dylib`dispatch_apply_f + 267
frame #9: 0x0000000100c9b447 libopencv_core.2.4.7.dylib`cv::parallel_for_(cv::Range const&, cv::ParallelLoopBody const&, double) + 117
frame #10: 0x0000000100705dc2 libopencv_imgproc.2.4.7.dylib`cv::remap(cv::_InputArray const&, cv::_OutputArray const&, cv::_InputArray const&, cv::_InputArray const&, int, int, cv::Scalar_<double> const&) + 2221
frame #11: 0x0000000100730153 libopencv_imgproc.2.4.7.dylib`cv::warpPerspectiveInvoker::operator()(cv::Range const&) const + 2341
frame #12: 0x0000000100c9b4d4 libopencv_core.2.4.7.dylib`(anonymous namespace)::block_function(void*, unsigned long) + 120
frame #13: 0x00007fff8fc308f8 libdispatch.dylib`_dispatch_client_callout2 + 8
frame #14: 0x00007fff8fc30849 libdispatch.dylib`_dispatch_apply_invoke + 87
frame #15: 0x00007fff8fc292ad libdispatch.dylib`_dispatch_client_callout + 8
frame #16: 0x00007fff8fc2b09e libdispatch.dylib`_dispatch_root_queue_drain + 326
frame #17: 0x00007fff8fc2c193 libdispatch.dylib`_dispatch_worker_thread2 + 40
frame #18: 0x00007fff8c8c3ef8 libsystem_pthread.dylib`_pthread_wqthread + 314
frame #19: 0x00007fff8c8c6fb9 libsystem_pthread.dylib`start_wqthread + 13
(lldb)
</pre>
```
## History
##### Ilya Lavrenov on 2014-02-11 09:36
```
- Category changed from calibration, 3d to imgproc, video
```
##### Kirill Kornyakov on 2014-02-17 07:50
```
Matthieu, could you please attach your sample here? Also, can you reproduce the issue with a smaller size image? It would be great to have access to it. May be you could upload it to some hosting?
Also, it would be great if you could debug the issue and send the fix.
- Description changed from Hi, I'm loading a very large image
(~600mb), then I use detector+descriptors... to Hi, I'm loading a
very large image (~600mb), then I use detector+descriptors... More
```
##### Kirill Kornyakov on 2014-02-17 07:54
```
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,category: imgproc,category: video,affected: 2.4 | low | Critical |
97,430,493 | opencv | compilation of objective-c code on MacOSx when using gcc compiler | Transferred from http://code.opencv.org/issues/3546
```
|| nicolas martin on 2014-02-15 04:01
|| Priority: Normal
|| Affected: branch 'master' (3.0-dev)
|| Category: build/install
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Mac OSX
```
## compilation of objective-c code on MacOSx when using gcc compiler
```
Hi,
On MacOsX platforms, the only check that's done for some files (e.g cap_qt.mm or window_cocoa.mm) is if APPLE is defined in CMakeLists.txt.
However, if the compiler is not clang, these files will fail to compile.
The compiler should also be checked.
```
## History
##### Kirill Kornyakov on 2014-02-17 07:14
```
Nicolas, is it possible to get a pull request from you?
Roman, do you think you're the right person to review it (or work on it)?
- Priority changed from High to Normal
- Assignee set to nicolas martin
- Category set to build/install
```
##### Kirill Kornyakov on 2014-02-17 07:54
```
- Status changed from New to Open
```
##### nicolas martin on 2014-02-21 21:01
```
Hi,
I pulled a request for this.
Thanks
```
| bug,auto-transferred,priority: normal,category: build/install,affected: 3.4 | low | Critical |
97,430,550 | opencv | (Java) Conversion between Mat to IplImage or BufferedImage | Transferred from http://code.opencv.org/issues/3548
```
|| Ernesto Moyano on 2014-02-15 14:58
|| Priority: Normal
|| Affected: None
|| Category: java bindings
|| Tracker: Feature
|| Difficulty: Medium
|| PR:
|| Platform: None / None
```
## (Java) Conversion between Mat to IplImage or BufferedImage
```
There's no function to convert the Mat type to swing compatible type. I already been working on a project that needs capture an 1920x1080 image from a webcam an display it into a swing container like a label or jpanel, but Mat type isn't compatible. So i try many ways to to this tasks but take to much time, but with this method i reduced the conversion time:
<pre>
public static BufferedImage matToBufferedImage(Mat mat) {
if (mat.height() > 0 && mat.width() > 0) {
BufferedImage image = new BufferedImage(mat.width(), mat.height(), BufferedImage.TYPE_3BYTE_BGR);
WritableRaster raster = image.getRaster();
DataBufferByte dataBuffer = (DataBufferByte) raster.getDataBuffer();
byte[] data = dataBuffer.getData();
mat.get(0, 0, data);
return image;
}
return null;
}
</pre>
would be cool a native code for doing this task.
```
## History
##### Kirill Kornyakov on 2014-02-17 07:58
```
Alexander, let's look at it.
- Assignee changed from Andrey Pavlenko to Alexander Smorkalov
- Status changed from New to Open
```
##### Alexander Smorkalov on 2014-03-16 15:47
```
- Description changed from There's no function to convert the Mat type
to swing compatible type. I alrea... to There's no function to
convert the Mat type to swing compatible type. I alrea... More
```
##### Alexander Smorkalov on 2014-04-02 01:28
```
- Target version changed from 2.4.9 to 3.0
```
##### Andrey Pavlenko on 2014-04-02 07:09
```
- Assignee changed from Alexander Smorkalov to Andrey Pavlenko
```
| auto-transferred,priority: normal,feature,category: java bindings | low | Minor |
97,430,783 | opencv | Java: important BackgroundSubtractionMOG2 parameters cannot be set | Transferred from http://code.opencv.org/issues/3573
```
|| Will Stewart on 2014-02-25 16:07
|| Priority: Normal
|| Affected: 2.4.8 (latest release)
|| Category: java bindings
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x86 / Linux
```
## Java: important BackgroundSubtractionMOG2 parameters cannot be set
```
After instantiating a BackgroundSubtractionMOG2, via parent class Algorithm we are able to getDouble() the parameters fTau and nShadowDetection, but setting does nothing for the fTau and creates the following error for nShadowDetection;
OpenCV Error: Bad argument (Argument error: the setter method was called for the parameter 'nShadowDetection' of the algorithm 'BackgroundSubtractor.MOG2', the parameter has unsigned char type, so it should be set by integer, unsigned integer, uint64, unsigned char, boolean, float or double value, but the setter was called with double value) in set, file /home/will/opencv-2.4.8/modules/core/src/algorithm.cpp, line 654
For example, the following code;
<pre>
BSMOG2 = new BackgroundSubtractorMOG2(2, 16, true);
double x = BSMOG2.getDouble("fTau");
System.out.println("BSMOG2.getDouble(fTau) = " + x);
double yy = BSMOG2.getDouble("nShadowDetection");
System.out.println("BSMOG2.getDouble(nShadowDetection) = " + yy);
</pre>
and the output is;
_BSMOG2.getDouble(fTau) = 0.5
BSMOG2.getDouble(nShadowDetection) = 127.0_
to set the values, I did;
<pre>
BSMOG2.setDouble("fTau", 0.7);
</pre>
EDIT: This executes without warning or complaint, but never really sets fTau, the value continues to be 0.5
And setDouble for nShadowDetection crashes with the following very odd statement;
OpenCV Error: Bad argument (Argument error: the setter method was called for the parameter 'nShadowDetection' of the algorithm 'BackgroundSubtractor.MOG2', the parameter has unsigned char type, so it should be set by integer, unsigned integer, uint64, unsigned char, boolean, float or double value, but the setter was called with double value) in set, file /home/will/opencv-2.4.8/modules/core/src/algorithm.cpp, line 654
On top of that, I tried other operations (paramType) to see what they resulted in;
_
Paramtype for fTau is 7
Paramtype for nShadowDetection is 11_
I can't figure out what these mean.
I also tried String paramhelpfTau = BSMOG2.paramHelp("fTau");
But that was _null_.
The BackgroundSubtractionMOG2 is virtually unusable if one cannot change the parameters.
```
## History
##### Will Stewart on 2014-02-27 15:28
```
This was my first bug report - did I do this right, or is there some other reason why it has been passed over for now?
```
##### Dinar Ahmatnurov on 2014-03-05 17:15
```
Alexander, please have a look
- Assignee set to Alexander Smorkalov
- Status changed from New to Open
- Category set to java bindings
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: java bindings | low | Critical |
97,430,855 | opencv | BestOf2NearestMatcher gives different result on second run | Transferred from http://code.opencv.org/issues/3579
```
|| Christopher Herbon on 2014-02-27 01:59
|| Priority: Normal
|| Affected: 2.4.8 (latest release)
|| Category: stitching
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Mac OSX
```
## BestOf2NearestMatcher gives different result on second run
```
When I excecute the folloing code twice with the exact same inputs, I get two different but similar results.
BestOf2NearestMatcher matcher(false, 0.3);
vector<MatchesInfo> temp_pairwise_matches = [...];
vector<ImageFeatures> temp_features = [...];
matcher(temp_features, pairwise_matches);
matcher.collectGarbage();
A little more info: I have spend a lot of time isolating the problem to this code. I create two different copies of BestOf2NearestMatcher and all input data is being cloned, so there is definitively no change in the input. The result on the first run is always the same and so is the result on the second and third run, and so on. But none of the results equals one of the previous results.
```
## History
##### Dinar Ahmatnurov on 2014-03-05 17:30
```
- Status changed from New to Open
```
##### Alexander Smorkalov on 2014-04-30 19:06
```
- Target version changed from 2.4.9 to 2.4.10
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: stitching | low | Critical |
97,430,896 | opencv | OpenCV 2.4.8: Possible code bug in frame blending in gpu/NPP_staging.cu | Transferred from http://code.opencv.org/issues/3580
```
|| Linchao Bao on 2014-02-27 03:03
|| Priority: Normal
|| Affected: 2.4.8 (latest release)
|| Category: gpu (cuda)
|| Tracker: Bug
|| Difficulty: Easy
|| PR:
|| Platform: x86 / Windows
```
## OpenCV 2.4.8: Possible code bug in frame blending in gpu/NPP_staging.cu
```
I was trying to figure out how is OpenCV doing the frame interpolation given forward and backward optical flow. I found the code in *modules/gpu/src/nvidia/NPP_staging/NPP_staging.cu* is very weird. Specifically, in function @nppiStInterpolateFrames@ (Line 2006~2097) it is first to warp the flow fields (to offset the reference pixel position for doing inverse warping, in my opinion). However, in the 4th call of function @nppiStVectorWarp_PSF2x2_32f_C1@, it is using @bwdU@ as the output buffer (Line 2079), where *I guess it should be @bwdV@ instead*. Checking the previous 3 call of the same function would easily discover that this should be a typo.
Another bug is inside function @BlendFramesKernel@, I believe Line 1936~1941, which is
@if (b0 && b1)
{
// pixel is visible on both frames
out[pos] = tex2D(tex_src0, x - _u * theta, y - _v * theta) * (1.0f - theta) +
tex2D(tex_src1, x + _u * (1.0f - theta), y + _v * (1.0f - theta)) * theta;
}@
*should be as follows instead* (because it is supposed to use bi-directional optical flow),
@if (b0 && b1)
{
// pixel is visible on both frames
out[pos] = tex2D(tex_src0, x - _u * theta, y - _v * theta) * (1.0f - theta) +
tex2D(tex_src1, x - _ur * (1.0f - theta), y - _vr * (1.0f - theta)) * theta;
}@
In practice, the two bugs are usually unnoticeable. But fixing the bugs does reduce artifacts in some cases. I suggest the authors check the code carefully and fix the obvious typo/logical errors. Thanks.
```
## History
##### Dinar Ahmatnurov on 2014-03-04 10:56
```
Vlad, please have a look
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: gpu/cuda (contrib) | low | Critical |
97,430,981 | opencv | iOS Installation Instructions are Out of Date | Transferred from http://code.opencv.org/issues/3587
```
|| Philip Salvaggio on 2014-03-04 02:25
|| Priority: Normal
|| Affected: 2.4.8 (latest release)
|| Category: documentation
|| Tracker: Bug
|| Difficulty: Easy
|| PR:
|| Platform: ARM / iOS
```
## iOS Installation Instructions are Out of Date
```
The instructions found on this page:
http://docs.opencv.org/doc/tutorials/introduction/ios_install/ios_install.html
no longer work for creating opencv2.framework. The symbolic link for /Developer is no longer needed. More importantly, the build_framework.py file must be executed from the opencv/platforms/ios/ directory. The new command should read.
cd ~/<my_working_directory>/opencv/platforms/ios/
python build_framework.py ~/<my_working_directory>/ios
```
## History
##### Dinar Ahmatnurov on 2014-03-04 10:53
```
Hi, Philip!
Could you please create a pull request for this? http://code.opencv.org/projects/opencv/wiki/How_to_contribute
- Assignee set to Roman Donchenko
- Status changed from New to Open
```
##### John Moses on 2014-03-14 05:32
```
Philip Salvaggio wrote:
> The instructions found on this page:
>
> http://docs.opencv.org/doc/tutorials/introduction/ios_install/ios_install.html
>
> no longer work for creating opencv2.framework. The symbolic link for /Developer is no longer needed. More importantly, the build_framework.py file must be executed from the opencv/platforms/ios/ directory. The new command should read.
>
> cd ~/<my_working_directory>/opencv/platforms/ios/
> python build_framework.py ~/<my_working_directory>/ios
I could not get either to work. I get the following error:
OSError: [Errno 2] No such file or directory: '../build/iPhoneOS-arm64/install/include/opencv2'
Additionally, whatever the solution is opencv/platforms/ios/readme.txt will need to be updated as well. Several other people are having similar problems:
http://stackoverflow.com/questions/16179120/error-compiling-the-opencv-framework
http://answers.opencv.org/question/14566/opencv-24-build-failed-error-while-creating-ios/
```
| bug,auto-transferred,priority: normal,category: documentation,affected: 2.4 | low | Critical |
97,431,052 | opencv | Opencv 2.4.8, MinGW, cv::CascadeClassifier(); | Transferred from http://code.opencv.org/issues/3593
```
|| Denis Gottardello on 2014-03-05 17:06
|| Priority: Normal
|| Affected: 2.4.8 (latest release)
|| Category: build/install
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x86 / Windows
```
## Opencv 2.4.8, MinGW, cv::CascadeClassifier();
```
In order to build OpenCV with MinGW32 is necessary to apply the patch described here:
http://stackoverflow.com/questions/21103042/error-while-building-opencv-monitorfromrect-was-not-declared-in-this-scope
But the serious problem is that the object *"cv::CascadeClassifier()"* is unusable.
The row
*"cv::CascadeClassifier *a= new cv::CascadeClassifier();"*
or
*"cv::CascadeClassifier a;"*
causes an application crash!
On Linux and Mac all works well.
```
## History
##### Denis Gottardello on 2014-03-05 17:07
```
The problem happens also with BackgroundSubtractorMOG2 and BackgroundSubtractorMOG.
```
##### Dinar Ahmatnurov on 2014-03-05 17:16
```
Roman. please have a look
- Priority changed from Blocker to Normal
- Assignee set to Roman Donchenko
- Category set to build/install
```
##### Dinar Ahmatnurov on 2014-03-05 17:17
```
- Status changed from New to Open
```
##### Denis Gottardello on 2014-06-17 19:49
```
Is this bug still open?
```
##### Denis Gottardello on 2015-01-01 17:58
```
Up!
```
| bug,auto-transferred,priority: normal,category: build/install,affected: 2.4 | low | Critical |
97,431,174 | opencv | Chamfer distance sample crashes with glibc double free or corruption | Transferred from http://code.opencv.org/issues/3602
```
|| Aleksandrs Ecins on 2014-03-11 17:01
|| Priority: Normal
|| Affected: 2.4.8 (latest release)
|| Category: legacy, contrib
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Linux
```
## Chamfer distance sample crashes with glibc double free or corruption
```
Hello,
Trying to run the sample Chmafer matching code from OpenCV (/trunk/opencv/samples/cpp/chamfer.cpp) I get the following error:
* glibc detected :double free or corruption (!prev): 0x0000000000745bb0 **
The reason is that in function @chamerMatching(img, tpl, results, costs)@ a template is created and then a reference to it is added inside @matcher_ object@:
<pre>
ChamferMatcher::Template template_(templ, (float)templScale);
ChamferMatcher::Matches match_instances = matcher_.matching(template_, img);
</pre>
As a result when the function returns @template_@ is destroyed first and then when @matcher_@ is destroyed it tries to destroy the same template again hence you gen an error.
I was able to get the code to run by commenting out the line that deletes templates in the desctructor of @Matching@:
<pre>
~Matching()
{
for (size_t i = 0; i<templates.size(); i++) {
//delete templates[i];
}
}
</pre>
However, is is not a solution to the proble since it can introduce memory leaks.
```
## History
##### Ivan Korolev on 2014-03-17 05:33
```
Hi Aleksandrs,
thanks for the bug report! Unfortunately, now we have limited human resources. So if you could investigate this issue and get back with some additional information or send a fixing pull request on GitHub (http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute), it would be much appreciated!
- Assignee changed from Vadim Pisarevsky to Aleksandrs Ecins
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,category: legacy,category: contrib,affected: 2.4 | low | Critical |
97,431,293 | opencv | java.lang.UnsupportedOperationException: Native object address is NULL in org.opencv.core.Mat.<init> | Transferred from http://code.opencv.org/issues/3613
```
|| Arsen L on 2014-03-16 19:24
|| Priority: Normal
|| Affected: 2.4.8 (latest release)
|| Category: java bindings
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Windows
```
## java.lang.UnsupportedOperationException: Native object address is NULL in org.opencv.core.Mat.<init>
```
Sporadically (quite rarely) receiving following exception in my java-application
<pre>
Exception in thread "pool-4-thread-4" java.lang.UnsupportedOperationException: Native object address is NULL
at org.opencv.core.Mat.<init>(Mat.java:424)
at org.opencv.utils.Converters.Mat_to_vector_Mat(Converters.java:266)
at org.opencv.utils.Converters.Mat_to_vector_vector_Point(Converters.java:500)
at org.opencv.imgproc.Imgproc.findContours(Imgproc.java:5851)
</pre>
Interesting, that if JVM launched in debug mode - this is not happen (issue relates to JVM?)
Tried on 1.7.0_45 and following one JVMs:
Java HotSpot(TM) 64-Bit Server VM; 1.7.0_51; 24.51-b03; mixed mode
```
## History
##### Daniil Osokin on 2014-03-17 07:11
```
Hi, Arsen!
Looks like it's an issue in your code. Anyway, please provide code sample, that reproduce this problem. Also may be you could debug this issue? It will help a lot!
- Assignee changed from Vadim Pisarevsky to Arsen L
```
##### Arsen L on 2014-03-19 10:25
```
Daniil,
I made working example (see file attached) on which you may reproduce this issue.
My current assumption is that this exception depends on some in-JVM concurrency optimizations.
I realize, that calling MyOpenCvUtils.init() in BorderDetectorWorker.borderField() smells, but leave it as is: it looks like it will increase probability to reproduce issue.
Replace path to image to real path (Main.IMAGE_PATH) on your machine. Also, launch JVM with -Djava.library.path="path_to_opencv_java248.dll" and declare library "opencv-248.jar" for sources which I prepared.
Please, try to launh Main.main() with Main.THREAS_NUM equals to number of your machine CPU cores.
After about 5-10 minutes of running you should see following excetion in console:
<pre>
Exception in thread "Thread-3" java.lang.UnsupportedOperationException: Native object address is NULL
at org.opencv.core.Mat.<init>(Mat.java:424)
at org.opencv.utils.Converters.Mat_to_vector_Mat(Converters.java:266)
at org.opencv.utils.Converters.Mat_to_vector_vector_Point(Converters.java:500)
at org.opencv.imgproc.Imgproc.findContours(Imgproc.java:5851)
at BorderDetectorWorker.borderField(BorderDetectorWorker.java:22)
at BorderDetectorWorker.run(BorderDetectorWorker.java:32)
at java.lang.Thread.run(Unknown Source)
</pre>
Note, that if laumced JVM in debug mode exception will not be thrown (so, there is no sense to debug code).
So, that's why I started to think this bug relates to Sun/Oracle JVM.
Please, let me know if you reproduced this issue or if I can assist you with something.
- File sources_to_reproduce.zip added
```
##### Arsen L on 2014-03-19 10:26
```
- Assignee changed from Arsen L to Daniil Osokin
```
##### Daniil Osokin on 2014-03-21 07:53
```
Thanks, Arsen. We will try to reproduce.
- Assignee changed from Daniil Osokin to Alexander Smorkalov
- Category changed from imgproc, video to java bindings
```
##### Sergei Nosov on 2014-03-24 15:07
```
- Status changed from New to Open
```
##### Piotr Betkier on 2014-09-08 12:41
```
I observe the same issue in my application when using OpenCV concurrently. I'm using OpenCV 2.4.9 on Linux.
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: java bindings | low | Critical |
97,431,342 | opencv | Inconsistent default parameters for SURF_GPU | Transferred from http://code.opencv.org/issues/3616
```
|| Nikolas Engelhard on 2014-03-19 09:35
|| Priority: Normal
|| Affected: 2.4.8 (latest release)
|| Category: gpu (cuda)
|| Tracker: Bug
|| Difficulty: Easy
|| PR:
|| Platform: Any / Any
```
## Inconsistent default parameters for SURF_GPU
```
Hello
I just started with OpenCV with Cuda support and wrote a little project using SURF_GPU and found a strange behaviour:
in GPU.hpp:
explicit SURF_GPU(double _hessianThreshold, int _nOctaves=4,
int _nOctaveLayers=2, *bool _extended=false*, float _keypointsRatio=0.01f, bool _upright = false);
but also
modules/nonfree/src/surf_gpu.cpp
cv::gpu::SURF_GPU::SURF_GPU()
{
hessianThreshold = 100;
*extended = 1;*
nOctaves = 4;
nOctaveLayers = 2;
keypointsRatio = 0.01f;
upright = false;
}
This code therefore fails:
SURF_GPU sgpu;
assert( ! sgpu.extended);
but this one is valid:
SURF_GPU sgpu(100);
assert( ! sgpu.extended);
Passing a value for the hessianThreshold should not change the descriptor size.
Fixing this is trivial but would also probably break some programs.
```
## History
##### Nikolas Engelhard on 2014-03-19 09:35
```
- Assignee deleted (Vladislav Vinogradov)
```
##### Daniil Osokin on 2014-03-20 09:17
```
Hi, thanks for the attention. Please, send a pull request with fix (http://code.opencv.org/projects/opencv/wiki/How_to_contribute).
- Description changed from Hello I just started with OpenCV with Cuda
support and wrote a little proje... to Hello I just started with
OpenCV with Cuda support and wrote a little projec... More
- Assignee set to Nikolas Engelhard
```
##### Nikolas Engelhard on 2014-03-20 10:50
```
Hello Daniil
I'll try to submit a patch but I won't have the time in the next weeks.
```
##### Daniil Osokin on 2014-03-21 07:49
```
Thanks, Nikolas! Looking forward to your pull request!
- Assignee changed from Nikolas Engelhard to Vladislav Vinogradov
```
##### Sergei Nosov on 2014-03-24 15:08
```
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: gpu/cuda (contrib) | low | Critical |
97,431,475 | opencv | Wrong types for constants in Matlab or wrong type conversion | Transferred from http://code.opencv.org/issues/3638
```
|| Yuval Ta on 2014-04-09 08:58
|| Priority: Normal
|| Affected: branch 'master' (3.0-dev)
|| Category: matlab bindings
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Windows
```
## Wrong types for constants in Matlab or wrong type conversion
```
I noticed this when calling cv.matchTemplate from matlab. The method type (cv.TM_CCOEFF_NORMED, ...) are of type 'double', and probably mistakenly converted to zero (thus always use the CV_TM_SQDIFF method) before calling the matchTemplate function.
The simple workaround is to convert the method type to 32 bit integer, but I assume it's not the intended way.
Doesn't work (return result of CV_TM_SQDIFF method):
cv.matchTemplate(I, T, cv.TM_CCOEFF_NORMED);
Works:
cv.matchTemplate(I, T, uint32(cv.TM_CCOEFF_NORMED));
This might be duplicate of #3465.
Thanks
```
## History
##### Dmitry Retinskiy on 2014-05-30 05:56
```
Hi Vadim,
could you check this problem?
Thanks.
```
##### Dmitry Retinskiy on 2014-05-30 05:56
```
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,affected: 3.4,category: matlab bindings | low | Critical |
97,431,562 | opencv | Python & .jp2 - imread() error on Windows only | Transferred from http://code.opencv.org/issues/3640
```
|| Yasuyuki Araki on 2014-04-10 08:30
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.8
|| Category: highgui-images
|| Tracker: Bug
|| Difficulty: Medium
|| PR:
|| Platform: x86 / Windows
```
## Python & .jp2 - imread() error on Windows only
```
Attempting to open some JPEG2000 files fails on Windows 7.
im = cv2.imread("chartsample.jp2")
error: cannot decode code stream
It succeeds on Linux (CentOS 6) and Mac (OS X 10).
Is there any workarounds?
```
## History
##### chuansheng wei on 2014-04-10 17:52
```
i've try myself to read the img on ubuntu and win7,it succeed on both system. maybe your windows7 system got some problem
```
##### Alexander Shishkov on 2014-05-06 20:34
```
- Affected version changed from 2.4.8 (latest release) to 2.4.0 -
2.4.8
- Difficulty set to Medium
- Assignee set to Alexander Smorkalov
- Category set to highgui-images
```
##### Andrew Senin on 2014-05-07 15:47
```
- Status changed from New to Open
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: imgcodecs | low | Critical |
97,431,655 | opencv | opencv2.4.x JpegDecoder::readData 'bool result = false' should be volatile modification | Transferred from http://code.opencv.org/issues/3649
```
|| Shine Sun on 2014-04-15 02:10
|| Priority: Normal
|| Affected: 2.4.0 - 2.4.8
|| Category: highgui-images
|| Tracker: Bug
|| Difficulty:
|| PR:
|| Platform: x64 / Linux
```
## opencv2.4.x JpegDecoder::readData 'bool result = false' should be volatile modification
```
opencv-2.4.8/modules/highgui/src/grfmt_jpeg.cpp:516
*bool result = false;*(bool JpegDecoder::readData( Mat& img )
should be
*volatile bool result = false;*
when opencv process some images which will cause "JpegDecoder::readData jpeg_finish_decompress" fail.
compiled in debug mode, readData returns true, but in some release compilation mode, readData returns false
like the program test.cpp:
using namespace std;
static jmp_buf buf;
void first(void) {
longjmp(buf,1);
printf("first\n");
}
int main() {
bool result = false;
if ( ! setjmp(buf) ) {
result = true;
first();
}
cout << result << endl;
return 0;
}
g++ test.cpp
it prints '1'
g++ test.cpp -O1
it prints '0'
g++ test.cpp -O2
it prints '0'
This is because the compiler to help us do the optimization, causes the value of result is modified.
So it must be with volatile, to prevent the value of the optimizing compiler.
When calling imdecode processing these pictures, will find that imdecode would fail, but imread was successful. As the above modified, return imdecode can, as we expected.
If you have any questions, you can contact me.
```
## History
##### Evgeny Talanin on 2014-04-16 07:39
```
Thanks for reporting.
You can propose this fix and a corresponding test as a pull request as described at http://code.opencv.org/projects/opencv/wiki/How_to_contribute page.
This ensures your changes to be tested using OpenCV's build system before affecting code in repository.
- Assignee set to Shine Sun
- Category set to highgui-images
```
##### Shine Sun on 2014-04-18 02:11
```
Hi Evgeny:
This is my first time using git,I find difficulty when submitted to the opencv code. When I use " git push" command, it returns erorr like this "fatal: unable to access 'https://github.com/Itseez/opencv.git/': The requested URL returned error: 403" . How to solve this problem.
```
##### Shine Sun on 2014-04-18 02:11
```
Hi Evgeny:
This is my first time using git,I find difficulty when submitted to the opencv code. When I use " git push" command, it returns erorr like this "fatal: unable to access 'https://github.com/Itseez/opencv.git/': The requested URL returned error: 403" . How to solve this problem.
Evgeny Talanin wrote:
> Thanks for reporting.
> You can propose this fix and a corresponding test as a pull request as described at http://code.opencv.org/projects/opencv/wiki/How_to_contribute page.
> This ensures your changes to be tested using OpenCV's build system before affecting code in repository.
```
##### Evgeny Talanin on 2014-04-18 10:00
```
Yes, that's a bit tricky at the first time.
That's an expected behavior: nobody has rights to push directly to OpenCV repository since only code tested by our build system should get there. It seems like you missed a concept of forking mentioned in 'The instruction in brief' section at http://code.opencv.org/projects/opencv/wiki/How_to_contribute.
The correct git workflow is:
1. Fork OpenCV repo under your account (see https://help.github.com/articles/fork-a-repo for details) and then only use your fork for implementing new features.
2. Push your changes to your OpenCV fork (not https://github.com/Itseez/opencv.git but https://github.com/YOUR_ACCOUNT_NAME/opencv.git)
3. Create pull request from your fork to OpenCV's repository (see https://help.github.com/articles/using-pull-requests for details).
4. Wait until pull request will be automatically tested and manually reviewed.
5. Enjoy your contribution to the open source computer vision library :)
If still having troubles, please just make a patch (or just a set of changed files) and attach here, we'll submit it to testing.
```
##### Shine Sun on 2014-04-22 01:49
```
Hi Evgeny:
Thank you for your answer, I tried, but without success, I think the patch to you to let you do a better patch addresses:.. Http://shinenuaa.oss-cn-hangzhou.aliyuncs.com/3649.patch Bug described in my above post. If you have any questions, please contact me, thank you!
Evgeny Talanin wrote:
> Yes, that's a bit tricky at the first time.
>
> That's an expected behavior: nobody has rights to push directly to OpenCV repository since only code tested by our build system should get there. It seems like you missed a concept of forking mentioned in 'The instruction in brief' section at http://code.opencv.org/projects/opencv/wiki/How_to_contribute.
>
> The correct git workflow is:
> 1. Fork OpenCV repo under your account (see https://help.github.com/articles/fork-a-repo for details) and then only use your fork for implementing new features.
> 2. Push your changes to your OpenCV fork (not https://github.com/Itseez/opencv.git but https://github.com/YOUR_ACCOUNT_NAME/opencv.git)
> 3. Create pull request from your fork to OpenCV's repository (see https://help.github.com/articles/using-pull-requests for details).
> 4. Wait until pull request will be automatically tested and manually reviewed.
> 5. Enjoy your contribution to the open source computer vision library :)
>
> If still having troubles, please just make a patch (or just a set of changed files) and attach here, we'll submit it to testing.
```
##### Shine Sun on 2014-04-24 02:18
```
- Assignee changed from Shine Sun to Evgeny Talanin
```
##### Dmitry Retinskiy on 2014-05-28 04:21
```
- Affected version changed from 2.4.8 (latest release) to 2.4.0 -
2.4.8
- Priority changed from High to Normal
- Assignee changed from Evgeny Talanin to Shine Sun
- Status changed from New to Open
```
##### Radek Svoboda on 2014-05-28 08:37
```
The other thing is that the return values of all jpeg_* function calls are ignored, so don't expect readData() to be safe in production software. I would recommend to also correct them at least for grfmt_jpeg.cpp in this Bugfix.
```
##### Shine Sun on 2014-06-05 11:56
```
thanks!
Radek Svoboda wrote:
> The other thing is that the return values of all jpeg_* function calls are ignored, so don't expect readData() to be safe in production software. I would recommend to also correct them at least for grfmt_jpeg.cpp in this Bugfix.
```
| bug,auto-transferred,priority: normal,affected: 2.4,category: imgcodecs | low | Critical |
97,431,728 | opencv | Bundle Adjuster in the Stitcher module is not converging with a high number of images | Transferred from http://code.opencv.org/issues/3657
```
|| Iacopo Masi on 2014-04-19 14:49
|| Priority: High
|| Affected: 2.4.8 (latest release)
|| Category: stitching
|| Tracker: Bug
|| Difficulty: Medium
|| PR:
|| Platform: x64 / Linux
```
## Bundle Adjuster in the Stitcher module is not converging with a high number of images
```
I found out that the *Bundle Adjuster module* in the recent *Stitcher module* is not converging if the software is provided with more than 20/30 images. The same bug is also reported from other users here [1]. The stitcher works for 5 to 10 images but not with a higher number. I found out that the problem is in the camera parameters refinement process (Bundle Adjuster). In particular I guess that the unbounded for loop [3] is never ending since the break sentence in this case is never reached. This only affect the camera parameters refinement module of the stitcher, the matcher and compositor work fine.
I guess that this break [2] in these cases where you have a high number of images >= 20 is _NEVER_ reached and if the module is forced to exit it reports nan in rotation matrices.
Please see an example attached to this ticket issue and a README that point on the dataset that has been used to fix the bug.
[1] http://answers.opencv.org/question/12713/nans-in-stitching_detailed-example-due-to-camera/
[2] https://github.com/Itseez/opencv/blob/master/modules/stitching/src/motion_estimators.cpp#L229
[3] https://github.com/Itseez/opencv/blob/master/modules/stitching/src/motion_estimators.cpp#L218
```
## History
##### Iacopo Masi on 2014-04-22 14:40
```
In particular today I discovered that the _err mat present in [4] after 3 or 4 iterations converges to a matrix full of nans (Not A Number) and this causes the infinite loop.
Iacopo Masi wrote:
> I found out that the *Bundle Adjuster module* in the recent *Stitcher module* is not converging if the software is provided with more than 20/30 images. The same bug is also reported from other users here [1]. The stitcher works for 5 to 10 images but not with a higher number. I found out that the problem is in the camera parameters refinement process (Bundle Adjuster). In particular I guess that the unbounded for loop [3] is never ending since the break sentence in this case is never reached. This only affect the camera parameters refinement module of the stitcher, the matcher and compositor work fine.
>
> I guess that this break [2] in these cases where you have a high number of images >= 20 is _NEVER_ reached and if the module is forced to exit it reports nan in rotation matrices.
> Please see an example attached to this ticket issue and a README that point on the dataset that has been used to fix the bug.
>
>
> [1] http://answers.opencv.org/question/12713/nans-in-stitching_detailed-example-due-to-camera/
> [2] https://github.com/Itseez/opencv/blob/master/modules/stitching/src/motion_estimators.cpp#L229
> [3] https://github.com/Itseez/opencv/blob/master/modules/stitching/src/motion_estimators.cpp#L218
[4] https://github.com/Itseez/opencv/blob/master/modules/stitching/src/motion_estimators.cpp#L222
```
##### Alexander Smorkalov on 2014-04-30 19:06
```
- Target version changed from 2.4.9 to 2.4.10
```
##### Rui Marques on 2014-06-11 11:51
```
Hello Iacopo Masi,
I am a bit confused by your last statement, you say "has been used to fix the bug.". Do you mean "to find the bug"?
I looked briefly at your code and seems like a test, not a fix, am I missing something?
```
| bug,auto-transferred,affected: 2.4,category: stitching | low | Critical |
Subsets and Splits