Steel Surface Defect Detection Based on YOLOv5s-CBC
In order to solve the problem of low detection accuracy of steel surface defects in industry,a steel surface defect detection method YOLOv5-CBC based on the improved YOLOv5s target detection net-work is proposed.Firstly,the CBAM attention mechanism is introduced in the backbone network to strengthen the feature extraction of steel defect images,and at the same time weaken the influence of the steel surface background on the detection results.Secondly,the nearest neighbor interpolation upsampling operator in the neck is replaced with the lightweight upsampling operator CAERAFE,which enhances the receptive field,improves the target detection ability and maintains light weight.Finally,the weighted bidi-rectional feature pyramid network BIFPN is used to modify the PANet in the original model to enhance the feature fusion ability of the model.The experimental results show that the average precision(mAP)of YOLOv5s-CBC on the NEU-DET dataset has reached 80.1%,which is 3.6%higher than the original YOLOv5s.This indicates that the method has good efficiency and detection accuracy,bringing an effective solution for the field of workpiece recognition.
万方数据
维普
