THE CLASSIFICATION AND RECOGNITION OF WRESTLING INJURY BASED ON DEEP LEARNING

• DOI: 10.1142/s0219519425400433
• Publication Date: 2025-03-13T09:57:56Z
• Authors: YANHUI ZHANG

This study proposes a deep learning (DL)-based multi-class injury classification and recognition model to improve the efficiency accuracy of diagnosis in wrestling. It also optimizes design by combining Convolutional Neural Networks (CNNs), residual networks, dense connection networks. Since existing diagnostic methods rely on manual experience is not fine enough, this designs framework automatically classify fractures, sprains, strains, other common wrestling injuries. The main problems include improving model’s accuracy, enhancing reasoning efficiency, optimizing computational complexity, solving data imbalance poor small sample categories. By comparing advanced models such as EfficientNet, Vision Transformer (ViT), Swin Transformer, comprehensively evaluates performance many dimensions. encompasses specificity, Area Under Receiver Operating Characteristic Curve (AUC-ROC), average cross-entropy loss. Regarding proposed optimized competition, training, accidental injuries 0.907, 0.930, 0.890, respectively, higher than most comparative models. In terms specificity AUC-ROC, AUC-ROC value training reaches 0.958, showing excellent performance. addition, loss low 0.242 competition injury, that can maintain error under multiple types. Therefore, has made an important contribution technology based DL sports medicine field. Especially, it provides reference support for applying intelligent medical future.