Steel surface defect detection algorithm based on TCS-YOLOv7

In order to improve the accuracy of steel surface defect detection, the TCS-YOLOv7 algorithm is proposed. Firstly, the Cycle GAN image enhancement technology was introduced, and the images were re-calibrated to increase the number of samples in the dataset. Secondly, the Triplet Attention mechanism is introduced into the backbone network of YOLOv7 to form the TAELAN module to improve the feature extraction ability of the network. At the same time, the SPPCSPCSE module is introduced in the neck, and the lightweight operator CARAFE is used to improve the efficiency of feature extraction and feature fusion. Then, the Parameter-Free Attention Mechanism (SimAM) was introduced in the head, and the module deduced the three-dimensional feature weights in the feature map, which improved the network's ability to learn features. Finally, the joint loss function Focal DIoU loss is used to further improve the accuracy of detection. Experimental results on the enhanced NEU-DET dataset show that compared with the YOLOv7 model, the accuracy of the TCS-YOLOv7 algorithm is increased by 3.7%, the recall rate is increased by 7%, the mAP@.5 value is increased by 4.1%, and the mAP value of the TCS-YOLOv7 algorithm reaches 80.9%. Therefore, the TCS-YOLOv7 algorithm is more suitable for complex industrial inspection scenarios.