In order to improve the accuracy and deployability of fatigue driving detection method, a fatigue driving detection algorithm based on RepVGG is proposed. An atrous spatial pyramid pooling (ASPP) module was added to the model to capture the multi-scale fatigue characteristics. A Convolutional block attention module (convolutional block attention module, CBAM) is combined with an ASPP module and separately applied to the model to further emphasize and capture the multi-scale information and important regional information expressed by fatigue features, and to suppress the background information in the image. Thereby improving the performance and robustness of the model. The results show that the accuracy of the improved RepVGG algorithm on the fatigue driving data set reaches 97. 34%, which is 2. 51% higher than that of the original algorithm, and the number of model parameters is only 7. 1 × 106, which has good detection accuracy and deployability.