| DefYOLOv12: A novel object detection model for detecting defects in Solar panels |
| Paper ID : 1026-ICEEM2025 (R2) |
| Authors |
|
Ali Mohammed Elhenidy *1, Saif M. Alsabti2, Raid gaib3, Lara R. Al-Najjar3 11Computer Engineering and Control Systems Department. Faculty of Engineering, Mansoura University, Egypt.
2 Faculty of Engineering mansoura national university, Gamasa, Egypt. 2Technical College of Engineering, Al-Bayan University, Baghdad, Iraq 3Technical College of Engineering, Al-Bayan University, Baghdad, Iraq |
| Abstract |
| Solar panels are becoming more and more important for the grid's sustainable electricity generation as the demand for renewable energy sources keeps growing. Defects in solar panels, however, have the potential to cause system instability and drastically reduce energy output. Consequently, it is crucial to identify such flaws as soon as possible. This study presents DefYOLOv12, a revolutionary object identification algorithm created especially to detect solar panel irregularities with extreme precision. In contrast to more intricate models like YOLOv12m, which are computationally costly but have comparable accuracy levels, DefYOLOv12 performs better and more efficiently. It uses about one-ninth of the parameters used by YOLOv12m, and achieves a top mean Average Precision at 0.5 (mAP@0.5) of 75.8%. The benefits of using deformable convolutions, which improve feature localisation and model generalisation without adding architectural complexity, are demonstrated by the success of DefYOLOv12. All things considered, this study shows that DefYOLOv12 provides a practical and efficient AI-based automated solar panel flaw detection solution. |
| Keywords |
| YOLOv12, solar panel inspection, defect detection, object detection, renewable energy |
| Status: Accepted |