10/32proof of conceptComputer Vision & Content Governance
Image Assurance
Checks property listing photos for quality, privacy and coverage, then a named reviewer decides.
What it solves
Listing photos can go out with poor quality, missing rooms or privacy risks. This checks every photo, and the gallery as a set, before a named reviewer approves it.
How it runs
Seven steps, input to output
Goes inThe photos for one property listing
- 01
Sign in
A reviewer signs in by name and work email, and the session is signed.
- 02
Upload
The originals are retained and never overwritten.
- 03
Technical gate
Deterministic rules check format, resolution and orientation.
- 04
Vision audit
A vision model classifies each scene and flags quality, privacy and misrepresentation risks.
- 05
Coverage
The gallery is judged as a set: every required room and the exterior must be covered.
- 06
Enhance
Exposure, contrast and edge-preserving sharpening only, shown beside the retained original.
- 07
Reviewer decision
A named reviewer approves or rejects every image, and the decisions export as calibration data.
Comes outAn approve or reject decision on every image, a readiness check for the listing, and calibration data
What it does
Governed AI review for property listing photography.
Deterministic rules check format, resolution and orientation, a vision model classifies each scene and flags quality, privacy and misrepresentation risks, and the gallery is judged as a set so a listing cannot be called ready until every required room and the exterior are covered.
Enhancement is deliberately narrow: exposure, contrast and edge-preserving sharpening only, always shown beside the retained original, with generative editing, object removal and virtual staging excluded by design.
A named reviewer approves or rejects every image, and those decisions export as calibration data.
Three decisions
What shaped the build
Enhancement is deliberately narrow
Exposure, contrast and edge-preserving sharpening only. Generative editing, object removal and virtual staging are excluded by design.
A listing is judged as a set
A listing cannot be called ready until every required room and the exterior are covered.
A named person makes the call
The model returns findings in a strict schema, and a named reviewer approves or rejects every image.
Built with
8 parts
- OpenAI Vision
- scene and risk audit
- Strict JSON schema
- structured findings
- fal.ai
- edge-preserving sharpen
- Rules engine
- format · resolution · size
- Coverage gate
- blocks incomplete listings
- Signed sessions
- named reviewer, 8h expiry
- Next.js · React
- review portal
- Vercel
- deploy target