Recognition of defects in wooden table spoons based on YOLO improved algorithm

Wooden tableware spoons form major defects such as mineral lines, cracks and scarring during the production process. In view of the current manual inspection is too inefficient as well as the problems of leakage and misdetection, so an improved algorithm based on YOLO for wood tableware defects recognition model WE-YOLO (Wee SE-YOLO) is proposed, which mainly solves the problem of detecting and distinguishing between mineral lines and cracks. The algorithm uses yolov5m, which is characterized by high detection accuracy, fast speed and small model, as a benchmark model. The training dataset is homemade, and the data are collected in five backgrounds, and the collected data are augmented; a layer of attention mechanism SE is added at the end of the backbone network of yolov5m, which improves the attention to different defects and improves the recognition accuracy; a layer of fine-target detecting header is added to the HEAD module, which reduces the probability of missing defects, and improves the detection accuracy of fine The detection accuracy of the defects is improved. Ablation experiments and comparison experiments of seven different models are carried out on the homemade dataset, and the improved algorithm has certain effects. The homemade dataset of this experiment alleviates the insufficiency of the existing dataset; the improved model is applicable to the two stages of wooden spoon production, which meets the requirements of industrial grade inspection standards.