CBFaceNet Micro Expression Recognition Network Based on Dual Attention Mechanism
Microexpression recognition plays an important role in student classrooms,healthcare,and other fields.Most existing micro expression recognition model technologies use traditional feature learning methods for feature extraction,but the recognition rate of traditional feature learning methods is not high,and deep learning methods will generate a large number of operating parameters.Therefore,a lightweight micro expression recognition method called CBFaceNet is proposed.This model can achieve end-to-end detection and is suitable for mobile devices with limited resources.In the proposed model,the fusion of three-dimensional attention mechanism simAM enhances the extraction of key features of micro expressions and can reduce model parameters.Insert channel and spatial attention module CBAM into the model to enrich the extracted facial features,and test the model using a mixed loss function.Comparing CBFaceNet with other models in the SMIC micro expression dataset,experimental results show that CBFaceNet has superior performance in recognition accuracy,complexity,and model parameters.
NETL
NSTL
万方数据
