Occluded Face Recognition Model Integrating Dynamic Mask, SCConv and CBAM

To solve the problem of insufficient face recognition accuracy caused by the lack of feature information in occluded scenes, this paper proposes a face recognition model based on the dynamic mask clearing mechanism and integrating the attention of SCConv and CBAM. The model empresses SCConv between the middle and upper layers of the pyramid extractor to extract more complete features. Meanwhile, a CBAM attention mechanism is introduced before the fully connected layer to further focus on the key areas. Experiments on the LFW, Occ-LFW, and Occ-WebFace datasets show that the accuracy rates of the model have reached 99.41%, 79.13%, and 73.84% respectively, outperforming the current mainstream occlusions face recognition models and significantly improving the accuracy rate of face recognition in occlusions scenarios.