Racial Faces in-the-Wild (RFW)


Racial Faces in-the-Wild (RFW) is a testing database for studying racial bias in face recognition. Four testing subsets, namely Caucasian, Asian, Indian and African, are constructed, and each contains about 3000 individuals with 6000 image pairs for face verification. They can be used to fairly evaluate and compare the recognition ability of the algorithm on different races.

The number of identities and images in RFW.


We construct our testing set similar to LFW. Besides,

•  RFW considers both the large intra-class variance and the tiny inter-class variance to avoid saturated performance.

•  Four testing subsets ensure to exclude other factors (e.g. pose, age and gender) except for race which can cause difference.


We examine some well-established methods, i.e. Center-loss, Sphereface, VGGFace2 and ArcFace, as well as four commercial recognition APIs, i.e. Face++, Baidu, Amazon, Microsoft on our testing set to study racial bias.

Face Verification Accuracy (%) on RFW dataset.


We provide loosely cropped faces, facial landmarks and lists of face pairs for testing:

  • Test.

  • Test Data_v1.
  • Test_lmk_v1. Estimated 5 facial landmarks on the provided loosely cropped faces.
  • Test_images_v1. The testing image list and label, e.g., ' m.0xnkj_ 0002.jpg 0'
  • Test_pairs_v1. 10 disjoint splits of image pairs are given, and each contains 300 positive pairs and 300 negative pairs similar to LFW.              
  •  Test_people_v1. The overlapped identities between RFW and MS-Celeb-1M and the number of images per identity.

The RFW dataset is available for non-commercial research purposes only. A complete version of the license can be found here. Permission to use but not reproduce or distribute the RFW database is granted to all researchers given that the following steps are properly followed:

Send an e-mail to Ms. Mei Wang (wangmei1@bupt.edu.cn) before downloading the database. You will need a password to access the files of this database. We will check the e-mail twice per week. Your Email MUST be set from a valid offical (University or Company) account and MUST include the following text:

 Subject: Application to download the RFW Face Database
 Name: (your first and last name)
 Affiliation: (University or Company where you work)
 Department: (your department)
 Position: (your job title)
 Email: (must be the email at the above mentioned institution)

 I have read and agree to the terms and conditions specified in the RFW face database webpage.
 This database will only be used for research purposes.
 I will not make any part of this database available to a third party.
 I'll not sell any part of this database or make any profit from its use.

If dropbox public links are temporarily suspended for generating excessive traffic, please try again several days later.

Note: there are overlaps between RFW and commonly used training dataset, i.e. MS-Celeb-1M, it is inconvenient to use it for training. So we remove their overlapping identities and release the remaining images of MS1M, i.e. MS1M_wo_RFW, for large-scale training.


Please cite the following if you make use of the dataset.

[1] Mei Wang, Weihong Deng, Jiani Hu, Xunqiang Tao, Yaohai Huang. Racial Faces in the Wild: Reducing Racial Bias by Information Maximization Adaptation Network. ICCV2019.

Bibtex  | PDF | Video

[2] Mei Wang, Yaobin Zhang, Weihong Deng. Meta Balanced Network for Fair Face Recognition. TPAMI 2021.

Bibtex  | PDF 

[3] Mei Wang, Weihong Deng. Mitigating Bias in Face Recognition using Skewness-Aware Reinforcement Learning. CVPR2020.

Bibtex  | PDF

[4] Mei Wang, Weihong Deng. Deep face recognition: A Survey. Neurocomputing.

Bibtex  | PDFVideo

Contact Us

Please contact Mei Wang and Weihong Deng for questions about the database.