Object Detection Algorithm Based on YOLO-PDE

This paper proposes an object detection algorithm based on YOLO-PDE and optimizes it for the oracle bone object detection scenario. This algorithm constructs a unique YOLOPDE network structure, including an input layer, multiple PDEBlock modules, SPPF modules and a detection layer. Inspired by partial differential equations, the PDEBlock module can enhance the feature extraction ability and effectively capture the complex stroke and texture features in oracle bone script. The SPPF module is used for spatial pyramid pooling, capable of processing input images of any size and fusing multi-scale information. Experiments on the Oracle dataset show that, compared with YOLO-C2f_Faster, the YOLO-PDE algorithm achieves higher mean average accuracy and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$F 1$</tex> value with fewer parameters and lower computing power consumption, demonstrating superior performance in the task of ancient character object detection. It provides efficient and accurate detection methods for the research and protection of oracle bone inscriptions.