3D / Point Cloud
Objaverse
800K+ annotated 3D models for training 3D and multimodal models.
100K–1M·CC-BY·Point Cloud (LAS/PLY)
GraspXL GraspXL is a large-scale dataset of physically plausible dexterous grasping motion sequences over Objaverse objects. It contains generated grasping motions for different hand models, together with processed object meshes used by the recorded sequences. The main archives cover diverse hand-object interactions where hands approach objects from different directions. The tabletop archives provide a complementary setting where objects are placed on a table and hands approach from… See the full description on the dataset page:
Source: Hugging Face Hub (ethHuiZhang/GraspXL). Metadata imported from the dataset’s Hub tags.