PCB Board Defect Detection Method Based on Knowledge Distillation
In order to enhance the efficiency of surface defect detection on PCB boards,a PCB board defect detection method based on knowledge distillation was proposed.Two network models were built:a large one with high accuracy but low efficiency,and a small one with high efficiency but lower accuracy.Using knowledge distillation method to distill the"knowledge"of the large model into the small model did not change the detection efficiency of the small model,but it would greatly improve the detection accuracy of the small model.The experimental results show that the average accuracy of the large model is as high as 0.938 while the intersection to union ratio is 0.5.After knowledge distillation.The detection efficiency of the small model on RTX4090 can reach a frame rate of 454.5.The proposed method has successfully achieved high-precision and efficient detection of defects in PCBs,which is of great significance.
万方数据
维普
