drprojects/superpoint_transformer
PyTorch implementation of transformer-based models for efficient 3D semantic and panoptic segmentation of point clouds.
Velocity · 7d
+0.9
★ / day
Trend
→steady
star history
This repository provides official implementations of Superpoint Transformer (ICCV'23), SuperCluster (3DV'24 Oral), and EZ-SP (ICRA'26) — deep learning methods for 3D point cloud segmentation. The approach uses transformer architectures and graph-based clustering to partition large-scale 3D point clouds into semantically and instance-level coherent superpoints for efficient panoptic segmentation.