基于改进YOLOv10s的木材表面缺陷检测模型OA北大核心

In response to the complex and diverse nature of wood surface defects and the real-time requirements of the detection tasks,this study aimed to improve the performance of the YOLOv10s model for wood surface defect detection.The proposed method simplifies the model structure by modifying the convolution modules in the Backbone,while introducing a lightweight dynamic upsampling module to enhance the sampling accuracy in the multi-scale feature fusion process.Additionally,a multi-scale attention module is incorporated to balance the detection of both large and small targets.Furthermore,a wood surface defect detection dataset was constructed to train and validate the model.Experimental results show that the proposed improved algorithm achieves an mAP@0.5(mAP@0.5 refers to the mean Average Precision at 50%intersection over union)of 95.1%and an inference time of 3.8 ms/frame.The results indicate that the improved model outperforms the current existing algorithms in detection accuracy,while it also meets the real-time detection requirements,making it suitable for wood surface defect detection tasks.