A Welding Defect Detection Method for Battery Pole Based on the Improved Yolov5

Welding defect detection plays an important role in the quality control of new energy batteries. Since the traditional manual detection methods are not intelligent enough and cost a lot, many deep learning algorithms have been proposed. With the development of detection technology, the Yolo series of algorithms have been applied to various detection tasks. Focus on our welding defect detection task for ithium battery's pole, an improved detection algorithm based on the Yolov5 model is proposed in this paper. Specifically, the SE attention module is added to the backbone, the last two C3 modules in the backbone are changed to ConvNeXtv2 modules, and the lightweight sampling operator CARAFE is introduced. The experimental results show that the improved algorithm increases the detection accuracy by 1.9% compared with the original Yolov5 model.