khaclinh commited on
Commit
e07c848
1 Parent(s): 9b125c7

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +207 -1
README.md CHANGED
@@ -1,3 +1,209 @@
1
  ---
2
- license: cc-by-sa-4.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ annotations_creators:
3
+ - expert-generated
4
+ language_creators:
5
+ - found
6
+ language:
7
+ - en
8
+ license:
9
+ - cc-by-nc-nd-4.0
10
+ multilinguality:
11
+ - monolingual
12
+ size_categories:
13
+ - 10K<n<100K
14
+ source_datasets:
15
+ - extended|other-wider
16
+ task_categories:
17
+ - object-detection
18
+ task_ids:
19
+ - face-detection
20
+ paperswithcode_id: wider-face-1
21
+ pretty_name: WIDER FACE
22
  ---
23
+
24
+ # Dataset Card for WIDER FACE
25
+
26
+ ## Table of Contents
27
+ - [Table of Contents](#table-of-contents)
28
+ - [Dataset Description](#dataset-description)
29
+ - [Dataset Summary](#dataset-summary)
30
+ - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
31
+ - [Languages](#languages)
32
+ - [Dataset Structure](#dataset-structure)
33
+ - [Data Instances](#data-instances)
34
+ - [Data Fields](#data-fields)
35
+ - [Data Splits](#data-splits)
36
+ - [Dataset Creation](#dataset-creation)
37
+ - [Curation Rationale](#curation-rationale)
38
+ - [Source Data](#source-data)
39
+ - [Annotations](#annotations)
40
+ - [Personal and Sensitive Information](#personal-and-sensitive-information)
41
+ - [Considerations for Using the Data](#considerations-for-using-the-data)
42
+ - [Social Impact of Dataset](#social-impact-of-dataset)
43
+ - [Discussion of Biases](#discussion-of-biases)
44
+ - [Other Known Limitations](#other-known-limitations)
45
+ - [Additional Information](#additional-information)
46
+ - [Dataset Curators](#dataset-curators)
47
+ - [Licensing Information](#licensing-information)
48
+ - [Citation Information](#citation-information)
49
+ - [Contributions](#contributions)
50
+
51
+ ## Dataset Description
52
+
53
+ - **Homepage:** http://shuoyang1213.me/WIDERFACE/index.html
54
+ - **Repository:**
55
+ - **Paper:** [WIDER FACE: A Face Detection Benchmark](https://arxiv.org/abs/1511.06523)
56
+ - **Leaderboard:** http://shuoyang1213.me/WIDERFACE/WiderFace_Results.html
57
+ - **Point of Contact:** shuoyang.1213@gmail.com
58
+
59
+ ### Dataset Summary
60
+
61
+ WIDER FACE dataset is a face detection benchmark dataset, of which images are
62
+ selected from the publicly available WIDER dataset. We choose 32,203 images and
63
+ label 393,703 faces with a high degree of variability in scale, pose and
64
+ occlusion as depicted in the sample images. WIDER FACE dataset is organized
65
+ based on 61 event classes. For each event class, we randomly select 40%/10%/50%
66
+ data as training, validation and testing sets. We adopt the same evaluation
67
+ metric employed in the PASCAL VOC dataset. Similar to MALF and Caltech datasets,
68
+ we do not release bounding box ground truth for the test images. Users are
69
+ required to submit final prediction files, which we shall proceed to evaluate.
70
+
71
+ ### Supported Tasks and Leaderboards
72
+
73
+ - `face-detection`: The dataset can be used to train a model for Face Detection. More information on evaluating the model's performance can be found [here](http://shuoyang1213.me/WIDERFACE/WiderFace_Results.html).
74
+
75
+ ### Languages
76
+
77
+ English
78
+
79
+ ## Dataset Structure
80
+
81
+ ### Data Instances
82
+
83
+ A data point comprises an image and its face annotations.
84
+
85
+ ```
86
+ {
87
+ 'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=1024x755 at 0x19FA12186D8>, 'faces': {
88
+ 'bbox': [
89
+ [178.0, 238.0, 55.0, 73.0],
90
+ [248.0, 235.0, 59.0, 73.0],
91
+ [363.0, 157.0, 59.0, 73.0],
92
+ [468.0, 153.0, 53.0, 72.0],
93
+ [629.0, 110.0, 56.0, 81.0],
94
+ [745.0, 138.0, 55.0, 77.0]
95
+ ],
96
+ 'blur': [2, 2, 2, 2, 2, 2],
97
+ 'expression': [0, 0, 0, 0, 0, 0],
98
+ 'illumination': [0, 0, 0, 0, 0, 0],
99
+ 'occlusion': [1, 2, 1, 2, 1, 2],
100
+ 'pose': [0, 0, 0, 0, 0, 0],
101
+ 'invalid': [False, False, False, False, False, False]
102
+ }
103
+ }
104
+ ```
105
+
106
+ ### Data Fields
107
+
108
+ - `image`: A `PIL.Image.Image` object containing the image. Note that when accessing the image column: `dataset[0]["image"]` the image file is automatically decoded. Decoding of a large number of image files might take a significant amount of time. Thus it is important to first query the sample index before the `"image"` column, *i.e.* `dataset[0]["image"]` should **always** be preferred over `dataset["image"][0]`
109
+ - `faces`: a dictionary of face attributes for the faces present on the image
110
+ - `bbox`: the bounding box of each face (in the [coco](https://albumentations.ai/docs/getting_started/bounding_boxes_augmentation/#coco) format)
111
+ - `blur`: the blur level of each face, with possible values including `clear` (0), `normal` (1) and `heavy`
112
+ - `expression`: the facial expression of each face, with possible values including `typical` (0) and `exaggerate` (1)
113
+ - `illumination`: the lightning condition of each face, with possible values including `normal` (0) and `exaggerate` (1)
114
+ - `occlusion`: the level of occlusion of each face, with possible values including `no` (0), `partial` (1) and `heavy` (2)
115
+ - `pose`: the pose of each face, with possible values including `typical` (0) and `atypical` (1)
116
+ - `invalid`: whether the image is valid or invalid.
117
+
118
+ ### Data Splits
119
+
120
+ The data is split into training, validation and testing set. WIDER FACE dataset is organized
121
+ based on 61 event classes. For each event class, 40%/10%/50%
122
+ data is randomly selected as training, validation and testing sets. The training set contains 12880 images, the validation set 3226 images and test set 16097 images.
123
+
124
+ ## Dataset Creation
125
+
126
+ ### Curation Rationale
127
+
128
+ The curators state that the current face detection datasets typically contain a few thousand faces, with limited variations in pose, scale, facial expression, occlusion, and background clutters,
129
+ making it difficult to assess for real world performance. They argue that the limitations of datasets have partially contributed to the failure of some algorithms in coping
130
+ with heavy occlusion, small scale, and atypical pose.
131
+
132
+ ### Source Data
133
+
134
+ #### Initial Data Collection and Normalization
135
+
136
+ WIDER FACE dataset is a subset of the WIDER dataset.
137
+ The images in WIDER were collected in the following three steps: 1) Event categories
138
+ were defined and chosen following the Large Scale Ontology for Multimedia (LSCOM) [22], which provides around 1000 concepts relevant to video event analysis. 2) Images
139
+ are retrieved using search engines like Google and Bing. For
140
+ each category, 1000-3000 images were collected. 3) The
141
+ data were cleaned by manually examining all the images
142
+ and filtering out images without human face. Then, similar
143
+ images in each event category were removed to ensure large
144
+ diversity in face appearance. A total of 32203 images are
145
+ eventually included in the WIDER FACE dataset.
146
+
147
+ #### Who are the source language producers?
148
+
149
+ The images are selected from publicly available WIDER dataset.
150
+
151
+ ### Annotations
152
+
153
+ #### Annotation process
154
+
155
+ The curators label the bounding boxes for all
156
+ the recognizable faces in the WIDER FACE dataset. The
157
+ bounding box is required to tightly contain the forehead,
158
+ chin, and cheek.. If a face is occluded, they still label it with a bounding box but with an estimation on the scale of occlusion. Similar to the PASCAL VOC dataset [6], they assign an ’Ignore’ flag to the face
159
+ which is very difficult to be recognized due to low resolution and small scale (10 pixels or less). After annotating
160
+ the face bounding boxes, they further annotate the following
161
+ attributes: pose (typical, atypical) and occlusion level (partial, heavy). Each annotation is labeled by one annotator
162
+ and cross-checked by two different people.
163
+
164
+ #### Who are the annotators?
165
+
166
+ Shuo Yang, Ping Luo, Chen Change Loy and Xiaoou Tang.
167
+
168
+ ### Personal and Sensitive Information
169
+
170
+ [More Information Needed]
171
+
172
+ ## Considerations for Using the Data
173
+
174
+ ### Social Impact of Dataset
175
+
176
+ [More Information Needed]
177
+
178
+ ### Discussion of Biases
179
+
180
+ [More Information Needed]
181
+
182
+ ### Other Known Limitations
183
+
184
+ [More Information Needed]
185
+
186
+ ## Additional Information
187
+
188
+ ### Dataset Curators
189
+
190
+ Shuo Yang, Ping Luo, Chen Change Loy and Xiaoou Tang
191
+
192
+ ### Licensing Information
193
+
194
+ [Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)](https://creativecommons.org/licenses/by-nc-nd/4.0/).
195
+
196
+ ### Citation Information
197
+
198
+ ```
199
+ @inproceedings{yang2016wider,
200
+ Author = {Yang, Shuo and Luo, Ping and Loy, Chen Change and Tang, Xiaoou},
201
+ Booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
202
+ Title = {WIDER FACE: A Face Detection Benchmark},
203
+ Year = {2016}}
204
+ ```
205
+
206
+ ### Contributions
207
+
208
+ Thanks to [@mariosasko](https://github.com/mariosasko) for adding this dataset.
209
+