Pointcept: Perceive the world with sparse points, a codebase for point cloud perception research. Latest works: Concerto (NeurIPS'25), Sonata (CVPR'25 Highlight), PTv3 (CVPR'24 Oral)

[NeurIPS'25] Official repository of Concerto: Joint 2D-3D Self-Supervised Learning Emerges Spatial Representations

[CVPR'24 Oral] Official repository of Point Transformer V3 (PTv3)

SAMPart3D: Segment Any Part in 3D Objects

[ECCV'24] OpenIns3D: Snap and Lookup for 3D Open-vocabulary Instance Segmentation

[CVPR'24 Highlight] GPT4Point: A Unified Framework for Point-Language Understanding and Generation.

[ICCV'23 Workshop] SAM3D: Segment Anything in 3D Scenes

[NeurIPS'22] An official PyTorch implementation of PTv2.