UNSEEN AI

UNSEEN AI is an AI-powered web application that investigates visible wall damage from photographs. Homeowners, tenants, and building caretakers often see peeling paint, stains, cracks, or damp-looking patches, but cannot tell whether the problem is minor surface damage or a sign of hidden moisture or leakage. Many people repaint over the damage without understanding its cause, allowing the underlying problem to worsen, while professional inspections can be costly or not immediately available. UNSEEN AI solves this problem by providing an accessible first visual assessment. A user captures a wall photo with a phone camera or uploads up to four photos. The app checks photo quality, sends the image through a FastAPI backend to Meta AI, and returns structured, evidence-based guidance: visible observations, uncertainties, possible causes with likelihoods, a confidence score, and safe next steps. It does not pretend to give a definitive diagnosis. Its key innovation is Next Best View. When evidence is insufficient, instead of guessing, the AI tells the user exactly which photo to take next — including the distance, angle, and the precise focus area marked on the image. Users can then continue a multi-round investigation with new photos and answers; the system compares the new evidence with the previous analysis, shows “What Changed,” and updates its confidence accordingly. Additional features include Automatic Damage Highlight bounding boxes, an interactive 3D wall visualization, camera capture guidance, PDF reports, saved case history, English/Sinhala language support, and an experimental Local AI mode that runs entirely in the browser for offline, private analysis. UNSEEN AI helps users detect wall problems earlier, document damage properly, avoid unsafe or unnecessary repairs, and make better-informed decisions about when to consult a qualified professional.