Expression Recognition in Complex Scenes Based on Fusion of VGGNet and ResNet

Building efficient and accurate expression recognition models in complex scenes is important for realizing sentiment analysis, psychology research and classroom management. To this end, we propose an efficient expression recognition model fusing VGGNet and ResNet, which is specifically designed to address the challenges of complex scenes such as lighting changes, occlusion, and pose changes. By using the two networks in parallel, combining the shallow feature extraction advantage of VGGNet and the deep feature extraction advantage of ResNet, and introducing the DCA module to enhance the feature fusion, the robustness and accuracy of the model are improved. The model achieves an average accuracy of 80.71% on the AffectNet dataset and 87.43% on the RAF-DB dataset, with a recall of 88.23% and an F1 value of 73.81% in the sample data. The experiments show that the method improves the accuracy of expression recognition in complex scenes while reducing the model complexity, which is a significant advantage over using VGGNet and ResNet alone.