Update README.md
Browse files
README.md
CHANGED
@@ -49,4 +49,76 @@ tags:
|
|
49 |
pretty_name: COCO2017-Colorization
|
50 |
size_categories:
|
51 |
- 100K<n<1M
|
52 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
49 |
pretty_name: COCO2017-Colorization
|
50 |
size_categories:
|
51 |
- 100K<n<1M
|
52 |
+
---
|
53 |
+
# COCO 2017 Dataset for Image Colorization
|
54 |
+
|
55 |
+
## Overview
|
56 |
+
|
57 |
+
This dataset is derived from the COCO (Common Objects in Context) 2017 dataset, which is a large-scale object detection, segmentation, and captioning dataset. The COCO 2017 dataset has been adapted here specifically for the task of image colorization.
|
58 |
+
|
59 |
+
## Dataset Description
|
60 |
+
|
61 |
+
- **Original Dataset:** [COCO 2017](https://cocodataset.org/)
|
62 |
+
- **Task:** Image Colorization
|
63 |
+
- **License:** [COCO dataset license](https://cocodataset.org/#termsofuse)
|
64 |
+
|
65 |
+
## Data origin
|
66 |
+
|
67 |
+
* Data originates from [cocodataset.org](http://images.cocodataset.org/annotations/annotations_trainval2017.zip)
|
68 |
+
|
69 |
+
## Format
|
70 |
+
|
71 |
+
```python
|
72 |
+
DatasetDict({
|
73 |
+
train: Dataset({
|
74 |
+
features: ['license', 'file_name', 'coco_url', 'height', 'width', 'date_captured', 'flickr_url', 'image_id', 'ids', 'captions'],
|
75 |
+
num_rows: 112268
|
76 |
+
})
|
77 |
+
validation: Dataset({
|
78 |
+
features: ['license', 'file_name', 'coco_url', 'height', 'width', 'date_captured', 'flickr_url', 'image_id', 'ids', 'captions'],
|
79 |
+
num_rows: 5000
|
80 |
+
})
|
81 |
+
})
|
82 |
+
```
|
83 |
+
|
84 |
+
## Usage
|
85 |
+
|
86 |
+
### Download image data and unzip
|
87 |
+
|
88 |
+
```bash
|
89 |
+
cd PATH_TO_IMAGE_FOLDER
|
90 |
+
|
91 |
+
wget http://images.cocodataset.org/zips/train2017.zip
|
92 |
+
wget http://images.cocodataset.org/zips/val2017.zip
|
93 |
+
|
94 |
+
unzip train2017.zip
|
95 |
+
unzip val2017.zip
|
96 |
+
```
|
97 |
+
|
98 |
+
### Loading the Dataset
|
99 |
+
|
100 |
+
You can load this dataset using the Hugging Face `datasets` library:
|
101 |
+
|
102 |
+
```python
|
103 |
+
from datasets import load_dataset
|
104 |
+
|
105 |
+
# Load the train split of the colorization dataset
|
106 |
+
train_dataset = load_dataset("nickpai/coco2017-colorization", split="train")
|
107 |
+
|
108 |
+
# Load the validation split of the colorization dataset
|
109 |
+
val_dataset = load_dataset("nickpai/coco2017-colorization", split="validation")
|
110 |
+
```
|
111 |
+
|
112 |
+
## Filtering Criteria
|
113 |
+
|
114 |
+
### 1. Grayscale Images
|
115 |
+
- Images in grayscale mode are identified and removed.
|
116 |
+
- Grayscale images lack color information and may not be suitable for certain tasks.
|
117 |
+
|
118 |
+
### 2. Identical Color Histograms
|
119 |
+
- Images with identical histograms across color channels (red, green, and blue) are removed.
|
120 |
+
- Such images may lack sufficient color variation, affecting model training and performance.
|
121 |
+
|
122 |
+
### 3. Low Color Variance
|
123 |
+
- Images with low color variance, determined by a specified threshold, are removed.
|
124 |
+
- Low color variance can indicate poor image quality or uniform color distribution.
|