AIFR: Face Recognition Research Based on Age Factor Characteristics

As we all know, the facial appearance change caused by age change leads to the low accuracy of face recognition, which is a significant difficulty in cross-age face recognition tasks. How to overcome the age problem, face feature extraction has become the key. This paper proposes a cross-age face recognition method based on deep learning. This method uses the Arcface loss function to realize cross-age face recognition by improving the residual neural network, combining it with the attention mechanism. Firstly, the face image is enhanced, and the Retinaface algorithm detects the face to complete the look preprocessing. Then the preprocessed face image is extracted by this model to achieve the purpose of cross-age face recognition. In addition, due to the lack of Asian face datasets in public data sets, this paper makes a self-use dataset based on the public data sets. It conducts experiments with FG-NET and CALFW datasets to confirm the universality of this method. The effect of the experimental training set reaches 92.67%, which makes other progress in cross-age face recognition.