About

An honest instrument, not an oracle

CivicLens inspects a physical miniature city — a tabletop model, a diorama, a school project — from a single captured frame. A vision model describes the scene, then reports the safety issues it can visually justify.

What it can detect

Road surface damage and potholes, vehicle collisions and possible accidents, riders without helmets, unsafe vehicle or pedestrian positions, blocked crossings, obstructions and visibly damaged infrastructure — as long as they are visible at model scale.

What it cannot do
  • See anything hidden, blurred, out of frame or too small to resolve.
  • Judge motion, speed, sound or anything outside a single still frame.
  • Replace a real inspection, engineering review or emergency service.
  • Guarantee a count — vehicle and figure totals are estimates.
Why every finding has evidence

Each finding states the visual reason it was flagged and a confidence value. Low-confidence observations are labelled uncertain rather than presented as fact, so you can check the model yourself and disagree.

Reliability

Two independent vision providers back the scan. If the primary analyser fails or returns unusable data, the scan switches to the backup and the report tells you it did.