Chinese, Japanese, and Korean Face Classification by Fusion of Multi-layers Channel Attention and Mutual Channel Loss

To solve the problem that the datasets are difficult to obtain and classify in the face classification task of young and middle-aged males in China, Japan, and Korea, the CJK_male face dataset was constructed first. Following the idea of fine-grained image classification, a classification method combining channel attention and mutual channel loss is proposed. The experimental results show that the accuracy of CJK_male classification reaches 88.26%, and it performs well when compared with other models of similar size or smaller.