Technology
A photo set goes through four stages before it reaches your team: ingestion, classification, scoring and feature detection, and structuring for delivery. Each stage is described below at the level of detail we'd walk through in a live technical review.
Pipeline
Raw photo sets are accepted in whatever order and orientation they arrive, with no requirement to shoot in a fixed sequence. Images are normalized for exposure and orientation before analysis, and grouped as a single property record.
Each image is classified by room type (kitchen, bedroom, bathroom, living space, exterior, and similar) with a confidence score attached to the label. Low-confidence classifications are flagged for a quick human check rather than silently guessed.
Each room is scored on a 1–10 condition scale, broken into sub-scores by area, alongside detection of notable features (renovation, natural light, outdoor space) and visible damage. Damage flags are rated by severity and mapped to the building system they relate to.
Results are compiled into a structured output record, an auto-ordered virtual tour sequence, and a draft floor plan layout, delivered in the formats your listing workflow already expects.
Output schema
| Field | Type | Description |
|---|---|---|
| room_type | string + confidence | Classified room label with a 0–1 confidence score |
| condition_score | number (1–10) + confidence | Overall condition score for the room, with confidence |
| sub_scores | object | Condition broken out by area: kitchen, bathroom, exterior, etc. |
| features | array | Detected features such as renovation, natural light, outdoor space |
| damage_flags | array of objects | Each flag includes severity and the building system it maps to |
| comparable_matches | array | Visually similar properties found in the same portfolio |
| tour_sequence | array | Auto-ordered image sequence for the virtual tour |
| floor_plan_draft | file reference | Draft floor plan layout generated from the photo set |
Integration
Structured output as JSON for direct integration, CSV export for spreadsheet-based workflows, and floor plan drafts in image and vector formats so they can be dropped into a marketing package or refined further.
Designed to sit alongside your existing MLS entry process and CRM, rather than replace either. Output is meant to populate fields your team already fills in by hand today.
We'll run a real photo set from one of your listings live, so you can see the pipeline stages against a property you already know.