Cross-age Face Image Similarity Measurement Based on Deep Learning Algorithms

In this study, a multi-feature fusion and decoupling solution based on the RNN is proposed from discriminative perspective. This method can address identity age information extraction losses in cross-age face recognition. not only constrains correlation between using loss but also optimizes feature restoration decoupling. The model was trained simulated CACD CACD-VS datasets. single-task learning stabilized after 125 iterations of training, while multi-task reached stable convergent state 75 iterations. terms performance analysis, DE-RNN had highest recognition accuracy with mAP 92.4%. Human Voting value 90.2%. Average 81.8%, whereas DAL lowest at 78.1%. Experiments proved that constructed study has effective application scenario.