<![CDATA[Solar modules play a crucial role in photovoltaic power generation systems, yet their performance can degrade due to physical and electrical damage. Therefore, automatic inspection is required to improve maintenance efficiency and prevent long-term performance loss. This study aims to implement an object detection approach for identifying solar module defects from visible RGB images using Faster R-CNN with a ResNet-50 backbone. The dataset was obtained from the Kaggle platform and manually annotated into PASCAL VOC format with two defect classes, namely physical damage and electrical damage, and expanded through data augmentation. The model was trained under several training configurations and evaluated using mean Average Precision (mAP), precision, recall, F1-score, and accuracy. The best performance was achieved using a batch size of 8, learning rate of 0.0001, and 30 epochs, resulting in 89% accuracy and 93% mAP. The results indicate that the model consistently detects both defect types and demonstrates potential for automated solar module inspection.]]>
Copyrights © 2025
