Discriminative Metric Learning with Convolutional Feature Descriptors for Age-Invariant Face Recognition and Verification

Aging includes internal and external factors that cause variation in appearance of face and, consequently, it is a difficult problem to handle in person identification and verification using face images. In this paper, we propose a method for face recognition and verification that is robust against variation of facial appearance caused by aging. Our proposed method uses discriminative metric learning over convolutional feature descriptors extracted from frontal face images. The results of an experiments for performance evaluation on the FG-Net and CACD face aging datasets empirically clarify that the proposed method is effective for improving the performance of person identification and verification in the scenario where input face images contain appearance variation due to aging.