Experimental analysis of the facial expression recognition of Male and female

With the development of deep learning, people have paid more and more attention to the research of facial expression recognition (FER), and obtained decent results in the laboratory. However, some studies have pointed out the defects of FER system itself based on the universal theory of expression and believed that human expression is specific. The purpose of this study is to analyze the influence of different gender data on the recognition rate of FER classification system. This study needs to prove that the recognition rate of different gender data in the existing FER system is different. In addition, it is necessary to confirm that there is a population recognition advantage between different gender groups in the experiment. Experiments construct a classification system by Inception V3 and transfer learning methods and design a comparative experiment. It was found that data sets with different gender ratios did influence the experimental results to some extent, and the recognition rate of female data was slightly higher than that of male data. Finally, it is concluded that models trained by male data have a higher rate of expression recognition for male group, as is the case with female data, which is similar to the situation of different cultural groups.