Facial recognition, one of the major research areas, has been adopted by organizations and governments for a few years now. Smartphone makers like Apple, Xiaomi, Huawei, Samsung, OPPO, Realme，among others, have been integrating this technology into their phones for providing maximum security to the users.
According to the study, facial recognition market size is expected to grow from USD 3.8 billion in 2020 to USD 8.5 billion by 2025, at a CAGR of 17.2% during the forecast period.
It takes a human 0.2 seconds to recognize a specific face，and most people can recognize about 5,000 faces. We also interpret facial expressions and detect emotions automatically. In other words, we’re naturally good at facial recognition and analysis.
In recent years, Computer Vision (CV) has been catching up and in some cases outperforming humans in facial recognition. Advancing CV and Machine Learning have created solutions that can handle tasks more efficiently and accurately than humans. (source：Forbes 20 January 2021.)
While there are so many databases in use currently, the choice of appropriate databases are so important that should be made based on the task given (emotions, aging, expressions, lighting etc).
In order to help researchers looking for the suitable datasets for their needs, we provide 9 datasets focused on human faces which are popular and high-quality. We’ll list some key characteristics and strengths and weaknesses of each.
1、CASIA-SURF HiFi Mask Data-set
Publication – SurfingTech
Released – 2021
Description – This dataset is CVPR 2021 Challenge contains both in-person reality videos and attack videos of each subject. There are totally 75 subjects (25 Asians, 25 Africans, 25 Caucasians). There is one-to-one correspondence between real person and its masks. Total is 62.4K videos
Main Use – Anti-spoofing face recognition
Size – 9T
Identities – 75
Data Gathering Method – Intel Realsense D435
2、Photo Attack Anti-spoofing Facial Dataset
Publication – SurfingTech
Released – 2019
Description – Photo Attack Anti-spoofing Facial Dataset is a large-scale face attributes dataset with 6K people faces, more than 48K videos.
Main Use – Anti-spoofing face recognition
Size – 10T
Identities – 6,000
Data Gathering Method – Intel Realsense SR300
3、Screen/Cloth Attack Anti-spoofing Facial Dataset
Publication – SurfingTech
Released – 2019
Description – Screen/Cloth Attack Anti-spoofing Facial Dataset is a large-scale face attributes dataset with 3K people faces, more than 42K videos.
Main Use – Anti-spoofing face recognition
Size – 15T
Identities – 3,000
Data Gathering Method – Intel Realsense SR300/D435i
4、3D Mask Attack Anti-spoofing Facial Dataset
Publication – SurfingTech
Released – 2019
Description – Screen/Cloth Attack Anti-spoofing Facial Dataset is a face attributes dataset with hundreds people faces, almost 6K videos.
Main Use – Anti-spoofing face recognition
Size – 3T
Identities – 148
Data Gathering Method – Intel Realsense D435i
5、2020 Anti-spoofing Facial Dataset
Publication – SurfingTech
Released – 2020
Description – 2020 Anti-spoofing Facial Dataset is a face attributes dataset with hundreds of people faces, almost 609.2K videos.
Main Use – Anti-spoofing face recognition
Size – 20T
Identities – 1,800
Data Gathering Method – Intel Realsense D435i
6、Multiracial 3D Multi-expression Facial Dataset
Publication – SurfingTech
Released – 2020
Description – Multiracial 3D Multi-expression Facial Dataset is a face attributes dataset with hundreds of people faces, almost 50.4K videos，and cover more than 20 countries and different ethnicity.
Main Use – Ai training
Size – 2T
Identities – 8,400
Data Gathering Method – Intel Realsense D435
7、3D body scan Dataset
Publication – SurfingTech
Released – 2021
Description – 3D body scan Dataset is a body attributes dataset with hundreds of people bodies, almost 36K videos，depth data.
Main Use – Ai training
Size – 2T
Identities – 300
Data Gathering Method – Intel Realsense D455
8、African 3D Multi-posture Facial data
Publication – SurfingTech
Released – 2020
Description – African 3D Multi-posture Facial data is a face attributes dataset with hundreds of people faces, almost 18K videos.
Main Use – Ai training
Size – 2T
Identities – 3,000
Data Gathering Method – Intel Realsense SR300
9、Chinese 3D HD Facial Dataset
Publication – SurfingTech
Released – 2020
Description – Chinese 3D HD Facial Dataset is a face attributes dataset with hundreds of super high-definition 3D face expressions data , almost 11K videos.
Main Use – Ai training
Size – 10T
Identities – 850
Data Gathering Method – 3DMD
10、South Asian 3D Multi-expression Facial Dataset
Publication – SurfingTech
Released – 2020
Description – South Asian 3D Multi-expression Facial Dataset is a face attributes dataset with hundreds of face expressions data, almost 12K videos.
Main Use – Ai training
Size – 2T
Identities – 2,000
Data Gathering Method – Intel Realsense SR300
