Defects of Injection Molding Parts Based on the Improved YOLOv7 Algorithm Research on Detection Technology

In view of the problem of manual visual inspection and difficulty in ensuring accuracy. Using YOLOv7 add SE attention mechanism module (Squeeze-And-Excitation) to replace the convolution layer, by improving the target detection algorithm of YOLOv7 to various defect image training, through the data shows that the improved model in accuracy (Precision) increased by 5.3%, recall rate (Recall) increased by 5.5 percentage points, the algorithm average accuracy mAP @ 0.5 increased by 8% the detection method can realize automatic detection, with high detection rate and detection accuracy.