Update README.md
Browse files
README.md
CHANGED
@@ -1,17 +1,17 @@
|
|
1 |
-
---
|
2 |
-
license: other
|
3 |
-
license_name: server-side-public-license
|
4 |
-
license_link: https://www.mongodb.com/licensing/server-side-public-license
|
5 |
-
task_categories:
|
6 |
-
- object-detection
|
7 |
-
- image-segmentation
|
8 |
-
tags:
|
9 |
-
- fashion
|
10 |
-
- e-commerce
|
11 |
-
- apparel
|
12 |
-
size_categories:
|
13 |
-
- 1K<n<10K
|
14 |
-
---
|
15 |
|
16 |
# FashionFail Dataset
|
17 |
|
@@ -20,6 +20,9 @@ The FashionFail dataset, proposed in the paper ["FashionFail: Addressing Failure
|
|
20 |
comprises 2,495 high-resolution images (2400x2400 pixels) of products found on e-commerce websites. The dataset is divided into training, validation, and test sets, consisting of 1,344, 150, and 1,001 images, respectively.
|
21 |
|
22 |
|
|
|
|
|
|
|
23 |
### Download Dataset
|
24 |
To address concerns regarding data regulations, we share only the URLs of the images, rather than sharing the image files directly.
|
25 |
However, we provide a simple script to facilitate dataset construction.
|
@@ -37,8 +40,7 @@ Then, execute the following script:
|
|
37 |
python fashionfail/data/make_dataset.py
|
38 |
```
|
39 |
which constructs the dataset inside `"~/.cache/fashionfail/"`.
|
40 |
-
An optional argument `--save_dir` can be set to construct the dataset in the preferred directory
|
41 |
-
it is not recommended to alter the default location.
|
42 |
|
43 |
### Annotation format
|
44 |
|
|
|
1 |
+
---
|
2 |
+
license: other
|
3 |
+
license_name: server-side-public-license
|
4 |
+
license_link: https://www.mongodb.com/licensing/server-side-public-license
|
5 |
+
task_categories:
|
6 |
+
- object-detection
|
7 |
+
- image-segmentation
|
8 |
+
tags:
|
9 |
+
- fashion
|
10 |
+
- e-commerce
|
11 |
+
- apparel
|
12 |
+
size_categories:
|
13 |
+
- 1K<n<10K
|
14 |
+
---
|
15 |
|
16 |
# FashionFail Dataset
|
17 |
|
|
|
20 |
comprises 2,495 high-resolution images (2400x2400 pixels) of products found on e-commerce websites. The dataset is divided into training, validation, and test sets, consisting of 1,344, 150, and 1,001 images, respectively.
|
21 |
|
22 |
|
23 |
+
> Note: The annotations are **automatically** generated by foundation models. However, a human annotator reviewed each sample to ensure the accuracy of the annotations.
|
24 |
+
|
25 |
+
|
26 |
### Download Dataset
|
27 |
To address concerns regarding data regulations, we share only the URLs of the images, rather than sharing the image files directly.
|
28 |
However, we provide a simple script to facilitate dataset construction.
|
|
|
40 |
python fashionfail/data/make_dataset.py
|
41 |
```
|
42 |
which constructs the dataset inside `"~/.cache/fashionfail/"`.
|
43 |
+
An optional argument `--save_dir` can be set to construct the dataset in the preferred directory.
|
|
|
44 |
|
45 |
### Annotation format
|
46 |
|