We're building a privately owned 3D point-cloud collection — so the community can own its own AI infrastructure instead of renting intelligence from a handful of corporations. Right now, we're gathering as many scans as we can.
A 3D scan is just millions of points in space — a colourless, structureless cloud. You can look at it, but nothing in the scan knows what it's looking at: a wall, a desk, the floor. 3DC adds that missing intelligence, labelling every point so the geometry finally means something.
3DC's core capability is instance segmentation — separating individual objects rather than just classifying surface types. It doesn't only see "furniture"; it sees this chair, that table, and the cabinet beside them as distinct, addressable things. Each object gets its own colour, its own boundary, its own identity — and that distinction is what makes a scan usable.
3DC ingests point clouds straight from Matterport scanners in E57 and PLY format. What comes back is the same geometry, now fully labelled — every point assigned to a surface type and an individual object instance, ready for whatever comes next.
Sitting on point clouds that no system can read? Affinity turns them into labelled, object-aware models ready for your pipeline.
Start the Conversation →