This article introduces an improvement in the YOLOv5 architecture by incorporating the CBAM (Convolutional Block Attention Module) attention module at the neck network end. CBAM is added after each concatenation operation to enhance the focus on small targets and optimize the fusion features in the neck. The role of CBAM is to strengthen the extraction of features by automatically ignoring irrelevant information, focusing on the fusion of crucial features, thereby improving the model's analytical capabilities for complex scenes. Experimental results indicate that the addition of the CBAM module successfully enhances the YOLOv5s model by highlighting key features and suppressing unimportant ones. This results in output feature maps containing more valuable information, significantly improving the accuracy of object detection. This improvement has shown positive effects in small object detection, feature fusion, and model speed.