YOLOv5 Small Object Detection Method based on Improved Deformable Convolution

In response to the problems of poor robustness and low accuracy of object detection algorithms in complex scenes, the YOLOv5 model is improved. This method introduces Deformable Convolutional Networks on the basis of the original YOLOv5 algorithm, which can adapt to the geometric changes of data and enhance the network's modeling ability for shape changes; Simultaneously introducing the attention mechanism Squeeze and Excitation reduces the number of parameters, improves the network's feature extraction capability, and enhances the useful information in the feature map; Finally, fully utilizing the GIOU loss function can more comprehensively reflect the similarity between bounding boxes, thereby improving the performance of the object detection model. The experimental results show that the method proposed in this paper has significantly improved the accuracy, recall, and detection precision of detecting graphics compared to the original model, thereby enhancing the accuracy of object detection.