Factory Cable Slot Box Defect Detection Method Based on Improved YOLOv8

At present, there are fewer research studies on the defect detection of factory cable slot box in the market, and the detection results for the defects of cable slot box have the problems of low average accuracy and difficult to detect the defects of small targets. Based on these problems, the YOLOv8-CMRF-TAM-MPD model based on improved YOLOv8 is proposed. First, in the backbone network, the feature extraction fusion module SPPF at the end is replaced with CMRF module. It can significantly improve the detection accuracy and speed during model training while effectively reducing the number of parameters of the model. Then, in the neck network part, some of the C2f modules therein are replaced with triple attention mechanism modules. This module enhances the feature representation of the model more effectively without significantly increasing the computational burden, improving the detection accuracy and robustness. Finally, the original CIoU loss function is replaced with the MPDIoU loss function, which uses a more flexible geometric metric to reduce the detection error and enhance the robustness and scene adaptation of the model. The experimental results show that the improved YOLOv8-CMRF-TAM-MPD model has an accuracy of 94.6% and a recall of 92.3%. mAP is 96.9%, which is an improvement of 1.3% compared to the benchmark model YOLOv8n. The amount of parameters was reduced by 11.2%, achieving the effect of a lightweight model. The model is able to detect cable slot box defects more effectively and accurately.