Masked Facial Expression Recognition with Attention Mechanism based on Mini_Xception Network

Facial expression recognition is of crucial importance in a variety of applications, including human-computer interaction and health monitoring. However, when there is occlusion, especially when the mouth region is occluded, conventional FER methods often suffer from significant performance degradation. To address this challenge, a novel FER model, Mini-Xception+CBAM, is proposed. This combines a lightweight network architecture with an attention mechanism to enhance performance in occlusion scenarios. The proposed approach utilises the Mini-Xception architecture and integrates a Convolutional Block Attention Module (CBAM) to enable the model to dynamically focus on key facial regions, such as the eyes and eyebrows, which are less likely to be occluded. Extensive experimentation on the FER2013, SMFER2013, and MFER2013 datasets demonstrates that the proposed model outperforms numerous state-of-the-art baseline models, including Mini-Xception, MobileNetV2, ResNet-18, and Vision Transformer, achieving The model achieved 90.4% accuracy on the FER2013, 89.2% on the SMFER2013, and 91.5% on the MFER2013, along with higher precision, recall, and F1 score. Furthermore, the model has been designed to be efficient, with a parameter count of 0.68M and a floating point operations count of 30.1M, making it ideal for deployment on devices with limited resources.