Point Transformer V3

Assigns a semantic class to every point in a 3D scan.

Field
Cross-domain
Method
Deep learning
Data it takes
point-cloud
License
MIT

PTv3 is a general point-cloud segmentation backbone, trained on indoor scene benchmarks rather than any particular scientific domain. It labels each point with a class from whatever scheme it is trained on.

Where the forestry models answer a fixed question about trees, this is the starting point for a segmentation problem that has no existing model — a quarry face, a built structure, a cave survey. It needs labelled examples for the classes you care about.

Upstream project →

Catalog entry last checked 2026-08-19. All models →

Run Point Transformer V3 on your data

The lab runs this on its own compute, in a container, with the inputs and parameters recorded alongside the result. Initial scoping conversations are free.

Contact the Lab