Automatic Detection Method for Clothing Collar Shape Combining Attention Mechanism and YOLOv5s
In order to solve the influence of external factors such as light,human body posture,environmental noise and shooting equipment on the accuracy of clothing collar shape detection,this paper proposed an automatic clothing collar shape detection method integrating attention mechanism and YOLOv5s.First,11 categories of cloth-ing collar data sets were constructed and labeled;then the original YOLOv5s model by changing the activation func-tion and introducing the attention mechanism to improve the accuracy of model was improved detection; finally,it carried out training,verification and test on the improved model.The experimental results show that choosing Fre-LU as the activation function and integrating the CBAM attention mechanism into the original YOLOv5s model has a better detection effect.After the test set test,the improved model mAP@0.5 value can reach 0.824,and can process 27.78 frames of images per second,both of which are better than the faster RCNN and SSD512 methods.Therefore,the method in this paper can complete the automatic detection task of clothing collar shape under com-plex background.
万方数据
维普
