Research on face recognition in coal mine scene based on improved ResNet

So far, facial recognition technology in general scenarios has become quite mature. However, it still faces numerous challenges in special situations and environments. In this article, we propose a recognition method based on an improved ResNet network for facial recognition in coal mining scenarios. Firstly, we optimize the ResNet network structure by adding a Bottleneck skip connection. We also introduce BN (Batch Normalization) layers and Dropout layers to address the issue of overfitting caused by deepening the network. Furthermore, we incorporate the CBAM (Convolutional Block Attention Module) attention mechanism into the feature extraction part of the ResNet network. By leveraging the spatial relationships within the feature maps, we enhance the weightage of facial texture features in key areas, capturing more facial contour information. The test results show that this optimized ResNet model exhibits improved accuracy in recognizing coal mine workers' faces in special scenarios. The proposed network structure is effectively applicable to facial recognition for coal mine workers, meeting the design requirements and possessing practical significance.