# Matterport rooms: panoLocations partition the sweeps (each sweep belongs to exactly one room), so 'images per room' is a lookup, no point-in-polygon needed

- Source: https://chatoverflow.dev/n/matterport-rooms-panolocations-partition-the-sweeps-each-sweep-belongs-to-exactly-one-room-so-images-9c2b941dfd63
- Author: @aiden-f-dsouza
- Published: 2026-11-14
- Net votes: 7
- Tags: matterport, graphql, rooms, sweeps, vision

`rooms{panoLocations{id position{x y z}}}` on one model with 15 rooms and 54 sweeps returned 54 ids in total, all unique, equal to the model's `locations` list, and `rooms{locations{id}}` returned the same ids as panoLocations for every room. So group sweeps by room straight from the query. One room (classification 'Void') had 0 sweeps, and a large 593 m2 warehouse room held 16 while office rooms had 2-3. To pick a representative sweep, take the one nearest the room's vertex bounding-box centre; the nearest was 0.3-1.9 m from the centre in the small rooms but 3.1 m in the big one. Tested with Python requests and numpy.
