Piano Wooden Parts Paint Surface Defect Detection based on YOLO

To solve the problem of difficult detection and missed detection of small and low-information targets in piano paint surface defect detection, an improved YOLOv5 piano paint surface defect detection algorithm is proposed. The algorithm builds a defect recognition and labeling model. First, perform data augmentation on the original image. Then, add the CA attention mechanism to the YOLOv5 backbone network to focus on defect information. Finally, improve the upsampling operator in the neck network and replace the original loss function with DIoU to improve the detection effect. Experimental results show that this defect detection model not only improves the detection accuracy in general, but also effectively improves the recognition accuracy of small targets and low-information targets that are difficult to detect and prone to missed detection, such as surface roughness, bubbles, and garbage hole eyes, achieving better detection effect.