Research on Progressive Multi-Granularity Distracted Driving Behaviors Classification Based on Attention Mechanism

The distracted driving behavior of drivers is the main factor causing road accidents, so it's particularly important to quickly identify and warn them. This article proposes an improved convolutional block attention module(CBAM), which focuses on deep optimization of CBAM: first, the maximum pooling and average pooling results are fused, and then the number of channels is adjusted through a convolutional layer to achieve more effective feature extraction. Secondly, clever dimensionality enhancement improvements were made to the multi-layer perceptron to further enhance the module's representation ability. Integrate the optimized CBAM into ResNet50 as a key component of its feature extraction process. At the same time, the idea of progressive training was borrowed to improve the algorithm's information acquisition ability. This method achieved an accuracy of 92.6% on the distracted driving dataset State Farm Distracted Driver Detection, which is better than other conventional methods. The results show that this method has good recognition ability for distracted driving.