CelebA-Spoof: Large-Scale Face Anti-Spoofing Dataset with Rich Annotations
Yuanhan Zhang, Zhenfei Yin, Yidong Li, Guojun Yin, Junjie Yan, Jing Shao and Ziwei Liu
In ECCV 2020.
Abstract: CelebA-Spoof is a large-scale face anti-spoofing dataset that has 625,537 images from 10,177 subjects, which includes 43 rich attributes on face, illumination,environment and spoof types. Live image selected from the CelebA dataset. We collect and annotate spoof images of CelebA-Spoof. Among 43 rich attributes, 40 attributes belong to Live images including all facial components and accessories such as skin, nose, eyes, eyebrows, lip, hair, hat, eyeglass. 3 attributes belong to spoof images including spoof types, environments and illumination conditions.CelebA-Spoof can be used to train and evaluate algorithms of face anti-spoofing.
[02/2021] The technical report of CelebA-Spoof Challenge 2020 is released on arXiv.
[08/2020] The CelebA-Spoof Challenge 2020 will start together with ECCV 2020 SenseHuman Workshop.
[08/2020] The CelebA-Spoof Challenge submission example is open!!
[08/2020] AENet_C,S,G model and its inference code for the intra-dataset test have been released.
We host CelebA-Spoof Challenge 2020 based on the CelebA-Spoof dataset. The challenge will officially start together with ECCV 2020 SenseHuman Workshop. Registration is now open. If you are interested in soliciting new ideas to advance the state of the art in real-world face anti-spoofing, we look forward to your participation! The CelebA-Spoof Challenge submission example is avaliable here.
Google Drive and Baidu Drive are available now!
- Google Drive: downloading link
- Baidu Drive: downloading link (password: 61fd)
AENet_C,S,G model and its inference code for the intra-dataset test have been released. Please see the intra_dataset_code for more details.
We hired 8 collectors to collect spoof data and another 2 annotators to refine labeling for all data. To improve the generalization and diversity of the dataset, as shown in Figure below, we define three collection dimensions with fine-grained quantities:
- Five Angles: All spoof type need to traverse all five types of angles including ''vertical'', ''down'', ''up'', ''forward'' and ''backward''. The angle of inclination is between [-30°, 30°].
- Four Shapes: There are a total of four shapes, ''normal'', ''inside'', ''outside'' and ''corner''.
- Four Sensors: We collected 24 popular devices with four types, ''PC'', ''camera'', ''tablet'' and ''phone''.
Besides the annotation of Live/Spoof, Existing face anti-spoofing only annotate the spoof type. To further comprehensively investigate face anti-spoofing tasks from various perspectives, in CelebA-Spoof, we annotate 43 different annotations. 40 types of Face Attribute defined in CelebA plus 3 attributes of face anti-spoofing, including Spoof Type, Illumination Condition, and Environment.
| 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Spoof Type | Live | Photo | Poster | A4 | Face Mask | Upper Body Mask | Region Mask | PC | Pad | Phone | 3D Mask |
| Illumination Condition | Live | Normal | Strong | Back | Dark | ||||||
| Environment | Live | Indoor | Ourdoor |
[0:40]: Face attribute labels # [40]: Spoof type label # [41]: Illumination condition label # [42]: Environment label # [43]: Live/Spoof label
Based on these rich attributes, we further propose a simple yet powerful multi-task framework, namely AENet. Through AENet,we conduct extensive experiments to explore the roles of semantic informationand geometric information in face anti-spoofing.
- CelebA dataset:
Ziwei Liu, Ping Luo, Xiaogang Wang and Xiaoou Tang, "Deep Learning Face Attributes in the Wild", in IEEE International Conference on Computer Vision (ICCV), 2015
- The CelebA-Spoof dataset is available for non-commercial research purposes only.
- You agree not to reproduce, duplicate, copy, sell, trade, resell or exploit for any commercial purposes, any portion of the images and any portion of derived data.
- You agree not to further copy, publish or distribute any portion of the CelebA-Spoof dataset. Except, for internal use at a single site within the same organization it is allowed to make copies of the dataset.
The use of this software is RESTRICTED to non-commercial research and educational purposes.
@inproceedings{CelebA-Spoof,
title={CelebA-Spoof: Large-Scale Face Anti-Spoofing Dataset with Rich Annotations},
author={Zhang, Yuanhan and Yin, Zhenfei and Li, Yidong and Yin, Guojun and Yan, Junjie and Shao, Jing and Liu, Ziwei},
booktitle={European Conference on Computer Vision (ECCV)},
year={2020}
}