Research on Human Subcutaneous Hard Lump Detection Algorithm Based on Improved YOLOv5
Aiming at the problems of low accuracy and slow computation speed of traditional subcutaneous hard lump detec-tion methods ,a subcutaneous hard lump target detection algorithm based on improved YOLOv5 is proposed. First ,the backbone network CSPDarknet53 of the YOLOv5 model is improved by introducing the Faster module to replace the C3 module. Second ,model pruning is utilized to mitigate the computational complexity of the whole model while ensuring its performance. Finally ,Wise-IoU is introduced to improve the regression performance of the network further. The experi-mental results show that the subcutaneous hard lump target detection algorithm based on the improved YOLOv5 improves the accuracy by 1.2％ and reduces the number of parameters by 77.7％ compared to the original YOLOv5. The whole algorithm is more lightweight ,effectively improving the algorithm's accuracy for detecting subcutaneous hard lumps ,reducing the number of computational parameters ,and enhancing the algorithm's running speed.
object detectionsubcutaneous hard lumptuberculin skin testdeep learningYOLOv5
万方数据
维普
