CASIA-Face-Africa: A Large-Scale African Face Image Database

Face recognition is a popular and well-studied area with wide applications in our society. However, racial bias had been proven to be inherent most State Of The Art (SOTA) face systems. Many investigative studies on algorithms have reported higher false positive rates of African subjects cohorts than the other cohorts. Lack large-scale image databases public domain one main restrictions studying problem recognition. To this end, we collect database namely CASIA-Face-Africa which contains 38,546 images 1,183 subjects. Multi-spectral cameras are utilized capture under various illumination settings. Demographic attributes facial expressions also carefully recorded. For landmark detection, each manually labeled 68 keypoints. A group evaluation protocols constructed according different applications, tasks, partitions scenarios. performances SOTA without re-training as baselines. proposed along its annotations, preliminary results form good benchmark study essential aspects biometrics for subjects, especially preprocessing, feature analysis matching, expression recognition, sex/age estimation, ethnic classification, generation, etc. can downloaded from website.