HSANet: Hybrid Self-Attention Network Recognition Facial Micro Plastic Method

Abstract

Due to the large changes in facial features, the correct recognition rate of the original face is low. In view of the phenomenon, this experiment proposed a hybrid self-attention block structure for rec-ognizing faces with facial features changes. For this reason, 26 kinds of micro-plastic surgery small sample image data sets were made by ourselves. Integrating self-attention into the bottleneck block of the residual network improves the ability of the hybrid self-attention block to capture the features of each region of the image. The experiment on the small sample micro-plastic data sets shows that the hybrid self-attention network proposed in this experiment has a higher correct recognition rate: 89.70%, the correct recognition rate increased by 2.65% compared with ResNet50, and the correct recognition rate of the hybrid selfattention model with improved connection increased by 1.12% compared with the hybrid self-attention model without improved connection, and the net-work performance was also improved.