Face Expression Recognition Based on Dynamic Sparse Attention and Residual Network

To enhance the accuracy and efficiency of facial expression emotion classification, we introduce an upgraded residual ResNet-based network model. We employ the CBAM mechanism to reconstruct the feature maps and enhance the discrimination of facial expressions. At the same time, we introduce a Bi-Level Routing Attention module BRA to retain fine-grained details in the feature maps. Rigorous experiments demonstrate that our method achieves impressive accuracy rates of 97.63% and 68.77% on the CK+ and Fer2013 datasets, respectively. The experimental results clearly demonstrate the effectiveness and superior performance of our proposed approach. This study provides a useful contribution to the development of the field of emotion classification and has potentially important impacts on various practical applications, such as emotion intelligent recognition systems and emotion analysis tools. Future research directions may involve extending our method to adapt to a broader range of application scenarios and pursuing performance improvements. In conclusion, our work provides new ideas for the application of deep learning models and is expected to have a profound impact in the field of artificial intelligence.