3D / Point Cloud
Objaverse
800K+ annotated 3D models for training 3D and multimodal models.
100K–1M·CC-BY·Point Cloud (LAS/PLY)
Lifting Unlabeled Internet-level Data for 3D Scene Understanding CVPR 2026 Annotated 3D scene data is scarce and expensive to acquire. SceneVerse++ demonstrates that carefully designed data engines can leverage web-unlabeled videos to automatically generate training data for 3D scene understanding. This release contains 1019 scenes from internet videos, each with 3D reconstruction, instance segmentation and video metadata. For more details, please refer to the paper.… See the full description on the dataset page:
Source: Hugging Face Hub (bigai/SceneVersepp). Metadata imported from the dataset’s Hub tags.