From a room or façade to a candidate address.
Find Region and Find Street estimate where a photograph was taken by reading the scene around the subject. Find Property does something different. It looks for the specific property in the frame, comparing your image against the property imagery available for a city you select, and returns the addresses that look most like it.
The image can come from outside or inside the building. For an exterior photograph, the comparison can draw on details such as:
- Rooflines
- Window arrangements
- Porches and entrances
- Brick, siding, and trim
- Landscaping and building geometry
For an interior photograph — the part of this capability with no equivalent outdoors — it can draw on details such as:
- Room layout
- Flooring and wall treatments
- Cabinets, counters, and fixtures
- Doors, windows, and trim
- Lighting, furniture, and other persistent visual features
This is why interior geolocation is worth treating as its own problem. A frame shot inside a room may hold no landmark, no signage, no view out of a window, and no metadata — nothing an outdoor model can reason from. It can still hold a great deal of property-specific information: a tile pattern, a stair rail, a countertop, the shape of a doorway. Those details are useless for guessing a region and very good for identifying one property among many.
No signage, no view outside, no metadata required. Everything the search has to work with is in the room.
Returned as candidate #3 of 20, with the property's own photographs attached to it.
The candidate confirmed above, photographed from outside. A façade shot of this property would be searched against the same imagery.
- Rooflines
- Window arrangements
- Porches and entrances
- Brick, siding, and trim
- Landscaping and building geometry
Either way in, the output is the same: ranked addresses with the photographs behind them.
A defined place to search.
Find Property will not run until you pick a supported city. Your image is compared against the property imagery currently available for that location, so Raven needs to know which place to search before it can search anything.
Coverage is built out city by city and is still expanding, which is why the list of supported cities lives in the region picker inside the app rather than on this page — it changes. It is also why this mode behaves differently from Find Street, which can be pointed at several regions at once. Find Property takes exactly one city per search.
If the city is already known from the case, you can go straight to Find Property. If it is not, narrow it first — with the case material you already have, or with Raven's outdoor geolocation capabilities — and then search for the specific property.
Ranked properties with the evidence behind each match.
- Candidate addresses: street, city, state, postcode, and country for each property returned.
- Coordinates: latitude and longitude, copyable, with the map zoomed to the property rather than the neighbourhood.
- Ranked similarity scores: each candidate is scored so the list can be ordered and worked through.
- The property photographs: the imagery behind each match, with the closest one flagged as the primary.
- Side-by-side review: your source image held against each candidate's photographs, one at a time.
- Review states: mark each candidate confirmed, possible, or dismissed as you work.
The photographs are the part that changes how this capability is used. Other modes hand you a location to corroborate somewhere else; here the corroborating material arrives with the answer. An investigator can put the source frame beside the candidate's imagery and see, feature by feature, whether the same tile, the same cabinetry, or the same roofline appears in both.
View product interface
Use it directly—or after narrowing the location.
There are two common ways into this capability.
- The city is known. Upload the interior or exterior image, pick the city, and search for the property directly.
- The location is unknown. Use the case context you already have, or Raven's outdoor geolocation modes, to settle on a likely city — then search for the specific property inside it.
The second path is a way of arriving at a city, not a required sequence. Find Region and Find Street read outdoor scenes; a photograph taken inside a room usually cannot be run through them at all, which is exactly the case where Find Property is the only capability with something to work from.
It also pairs with Verify Image for the same reason every other capability does: a generated photograph of a room or a house that never existed will still be compared against real ones.
Common questions.
Does Find Property work with interior photographs?
Yes. In supported cities, Find Property can compare room layouts, finishes, fixtures, furniture, and other persistent visual details against available property imagery to surface candidate addresses.
Does it also work with exterior photographs?
Yes. Exterior searches can use building geometry, rooflines, windows, entrances, materials, and landscaping to identify visually similar properties.
Do I need to know the city?
Yes. Find Property currently searches one supported city at a time. The city may come from existing case information or another Raven geolocation result.
Can I search multiple cities simultaneously?
Not currently. Find Property searches one selected city per search.
Is it production ready?
Find Property is currently a beta capability. Coverage and performance vary by city, and every result should be visually corroborated.
