SunScout
The Project
SunScout is an offline mobile app that uses on-device AI to detect solar panel faults from photos. I built it because manual rooftop inspection is costly, risky, and often misses small problems that reduce clean energy output. Users import drone or camera images, run analysis on their phone, and review fault labels and confidence scores, all without internet. I trained an EfficientNetB0 model on 1,575 images (94.29% accuracy) and built the app in Flutter with ONNX Runtime. The biggest challenges were similar-looking fault types and fitting accurate AI onto a phone. SunScout aims to make solar inspection safer, cheaper, and more accessible, especially in remote areas.
