Lightweight-Designed Defect Detection Model for Photovoltaic Cells

Photovoltaic (PV) cell defect detection is a critical part of PV cell production. In actual manufacture, limited by the computing power of the equipment, there are high requirements for the accuracy and lightweight of PV cell defect detection models. This paper proposes a lightweight PV cell defect detection model based on the YOLOv8n algorithm. Specifically, the ADown operator replaces traditional Downsampling to reduce computational complexity, the C3Ghost module substitutes for the C2f module to eliminate redundant parameters, and Slim-BiFPN streamlines feature fusion through efficient multi-scale integration. The model achieves a 0.85% mAP@50 increase, Detection accuracy improved by 4.8%, and Computational complexity (GFLOPs) reduced by 44%, parameter count decreased by 63% and compressed by 57%, offering an efficient solution for industrial deployment.