Fig. 1. MCLResNet network architecture

Fig. 2. Core part of the hidden layer.（a）Structure of BasicBlock；（b）Structure of BasicBlock 1.

Fig. 3. Overall network structure diagram

Fig. 4. （a）Ordinary convolution；（b）Grouped convolution.

Fig. 5. Channel shuffle

Fig. 6. Inverted residual structure

Fig. 7. Improved channel attention module

Fig. 8. Improved spatial attention module

Fig. 9. Samples of FER2013 and CK+ datasets

Fig. 10. Sample diagram of data enhancement

Fig. 11. （a）Relationship between the ascending coefficient and the accuracy rate；（b）Relationship between the ascending coefficient and the number of model parameters.