A biscit defect detection method based on improved YOLOv5

Abstract

Defect detection is of great importance to ensure the quality of biscuit production. An improved YOLOv5 biscuit detection algorithm is proposed for the problems of poor real-time and low accuracy of biscuit defect detection methods. First, the number of C3s in the backbone network is reduced, and then the depth-separable convolution is used instead of the normal convolution in the network to reduce the model parameters and computation and improve the detection speed. Secondly, the SE attention module is added to the feature extraction layer to enhance the feature extraction capability of the backbone network and improve the accuracy of biscuit defect detection. Finally, the EIOU loss function is introduced to accelerate the model convergence and accurate target localization. The improved algorithm is tested on the self-built biscuit dataset, and the experimental results show that: the detection accuracy of the proposed algorithm can reach 99.2%, and the detection speed is 67 frames/s, which can meet the actual production requirements.