Surface Defect Detection Algorithm for Air Filters Based on Improved YOLOv7

This paper proposes a sample augmentation method to solve the problem of small samples in the inspection of surface defects of filters, the method can generate large-scale samples while ensuring diversity and authenticity, making deep network training feasible. And an improved YOLOv7 defect detection algorithm is proposed by modifying the network structure and adding attention mechanisms, using Wise-IoU to replace the original loss function to increase the accuracy and maintain the speed of detecting surface defects on filters. The results showed that the improved model achieved an average precision of 97.55%, an improvement of 5.76% over the original YOLOv7 algorithm, with a frame rate of 46.5 frames/s.