Solar Panel Defect Detection via Convolutional Neural Networks
Göksel Gündüz
YTU Journal of Computer Vision & Robotics· Vol. 9 (9)· 13 June 2026
Abstract
We present a real-time defect detection system for photovoltaic panels using aerial thermographic imagery. A custom EfficientNet-B4 architecture trained on 12,000 annotated thermal images achieves 94.3% mAP across five defect categories (hotspot, bypass diode failure, cell crack, soiling, and delamination) while running at 15 fps on an edge GPU.
Keywords
solar paneldefect detectionconvolutional neural networkthermographyEfficientNet
How to cite
Göksel Gündüz (2026). Solar Panel Defect Detection via Convolutional Neural Networks. YTU Journal of Computer Vision & Robotics, 9(9)
© 2026 the author(s). This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International (CC BY 4.0) licence.