Note on FA-TestV1
CASIA Face Image Database for Testing Version 1.0 (or FA-TestV1) contains 4,000 color facial images of 1,000 subjects. The face images of FA-TestV1 were captured using Logitech USB camera in one session. The volunteers of FA-TestV1 include graduate students, workers, waiters, etc. Typical intra-class variations include illumination, pose, expression, eye-glasses, imaging distance, etc.
Both FA-TestV1 and CASIA-FaceV5 are part of a larger database. There is overlap in subjects between FA-TestV1 and CASIA-FaceV5 but no overlap in images. CASIA-FaceV5 is released to public domain for training and developing face recognition algorithms. In contrast, FA-TestV1 is a sequestered dataset for testing submitted algorithms so only facial regions are cropped from the original images to avoid possible use of non-face regions such as clothes for personal identification. All face images of FA-TestV1 are 16 bit color BMP files and the image resolution varies with the size of facial region in the 640*480 original image.
Fig. 1 Example face images in FA-TestV1