YOLOv5 Detection Algorithm of Steel Defects Based on Introducing Light Convolution Network and DIOU Function

Surface will produce cracks, rolling scale, patches, pitting surfaces, inclusions, scratches, etc. These defects will damage the physical structure of the steel and have a great impact on the quality of the steel. Aiming at the problem that the detection accuracy of steel surface defects needs to be improved, an improved algorithm based on YOLOv5 is proposed to identify the surface defects of steel. First, the improved lightweight ShuffleNet V2 is introduced to replace the convolution layer in some backbone networks, so as to enhance the feature extraction capability of the backbone networks; Then DIoU loss function is selected to replace GIoU loss function in YOLO to optimize the training model to improve the recognition accuracy of the network model. The improved algorithm has obvious improvement in accuracy P (Precision), recall R (Recall), and average precision MAP (Mean Average Precision), which verifies the feasibility of this method.