Researchers have introduced SolarFCD, a new large-scale dataset designed to improve the automated detection of faults in solar photovoltaic systems. The dataset combines RGB/Drone and Thermal Infrared images, totaling 4,435 images across four defect categories: healthy, surface obstruction, structural fault, and electrical fault. To establish a benchmark, sixteen classification architectures were trained and evaluated, with ResNet101V2 achieving the highest accuracy of 86.68%. The dataset, annotations, and baseline code are being made publicly available to foster further research in this area. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Provides a new benchmark dataset to advance AI-driven inspection of solar panels, potentially improving efficiency and reducing maintenance costs.
RANK_REASON This is a research paper introducing a new dataset and benchmark for a specific AI application.