MLCA-YOLO: improved yolov8 for solar cell defect detection

Electroluminescence images are commonly used in defect detection of solar cells and photovoltaic modules. Defect identification is very complicated due to complex background interference. Therefore, we propose a deep learning model that combines the C2f module of the YOLOv8 backbone with the lightweight attention module of Hybrid Local Channel Attention (MLCA) to improve the reliability and accuracy of defect detection. This model is a lightweight attention module that simultaneously considers channel information and spatial information in the image, as well as local and global features. The validity of MLCA-YOLO was verified by using the public data set of PVEL-AD. The experimental results show that MLCA-YOLO is superior to the existing defect detection methods in terms of accuracy and robustness, and the accuracy of mAP50 is increased to 92.3%.