Solar Panel Defect Detection using FOMO on Edge Impulse

A machine learning model is developed for detecting defects in solar panels, focusing on hotspot detection. The project was created on Edge Impulse with images of solar panels in normal and defective states. The trained model with appropriate training accuracy was then deployed on a smartphone for real-time defect detection, utilizing a camera to capture and process image frames, overlaying results on a smartphone screen for immediate inspection. The evaluation revealed an F1 score of 80.5% for hotspot detection with an accuracy of over 95% for the model-detecting hotspots. This demonstrates the model’s practical application and effectiveness for solar panel hotspot detection.