Deep Learning Based Solar Panel Structural and Electrical Damage Detection

Guaranteeing the quality of solar panels is crucial, with numerous inspection techniques and technologies suggested. Traditional techniques for visually inspecting solar panels have several limitations and often fail to ensure perfect results. Automated inspection of solar panels utilizing deep learning has become a powerful and efficient method for tackling quality control problems in the industrial sector. Thus, this study concentrates on identifying defects in solar panels through a tailored and lightweight model built on MobileNetv2 for classification and feature extraction, along with the application of the Grad-CAM algorithm to pinpoint the damaged region in the panel. This research primarily targets improving the performance of real-time solar panel defect detection by utilizing a dataset with six different categories of solar panel defects. Additionally, we conducted a comparative analysis between our proposed model and the other models discussed in the literature review section. Our classifier attained an accuracy of 96 %, and Grad-CAM demonstrated effective results for locating defects.