LitePT
Point cloud segmentation · Indoor RGB point cloud or outdoor LiDAR sweep
About LitePT
LitePT assigns semantic labels to points in a 3D scan. The Space provides checkpoints for indoor RGB scans and outdoor LiDAR sweeps, with side-by-side input and segmented previews.
Choose a checkpoint that matches the kind of data you supply. The output label array preserves input point order, allowing it to be used alongside the original scan. This workflow classifies existing geometry; it does not reconstruct a mesh from a photograph.
How to use LitePT
- Upload a supported point-cloud file or select a bundled example.
- Choose the appropriate indoor or outdoor checkpoint and select Segment.
- Inspect the color legend and segmented preview, then download the per-point labels.
What to know before you start
- The interface expects coordinates in meters with the vertical axis pointing up.
- Input features and scene type should match the selected checkpoint. The visible file types include PLY, BIN, NPY, and text-based point data.
The provider controls availability, sign-in, processing, and downloads. Save any result before reloading or closing the demo.
Does LitePT generate a new 3D model?
No. It predicts semantic classes for an existing point cloud and returns labels plus a visual preview.
