Combining deep learning with R-CDT for solar defect recognition

In response to the lack of publicly available datasets for solar panel defects and the challenges in applying general recognition models, this study presents an algorithm tailored for small-sample solar panel defect detection. The algorithm leverages the strengths of deep learning models and the R-CDT transformation, initially employing Faster R-CNN for defect localization and segmentation, followed by the application of the R-CDT algorithm for defect classification. This approach effectively mitigates human-induced factors and reduces the risk of overfitting in small-sample scenarios. Experimental results demonstrate its high effectiveness as a small-sample recognition method, surpassing popular models (Faster R-CNN and YOLOv5) in terms of recognition accuracy.