UNSW and DTU test SimpleResNet for PV module image enhancement

Researchers from the University of New South Wales, Sydney, and the Technical University of Denmark, Roskilde, have developed a deep learning method to denoise outdoor PV module images. Their model, SimpleResNet, improves the quality of electroluminescence images by preserving fine details. It was tested against conventional techniques and showed better speed and performance. The method is suitable for large-scale inspections. A complete preprocessing pipeline was also introduced. It corrects perspective distortion, removes noise, and enhances sharpness. The pipeline was validated using real outdoor images. It performed consistently across different module types and lighting conditions. The approach helps improve the accuracy and speed of automated PV module inspections.