3D / Point Cloud
Objaverse
800K+ annotated 3D models for training 3D and multimodal models.
100K–1M·CC-BY·Point Cloud (LAS/PLY)
GObjaverse-LVIS Dataset A curated subset of the GObjaverse-280K dataset filtered to objects present in the LVIS (Large Vocabulary Instance Segmentation) taxonomy and annotated with Cap3D text captions. Dataset Summary Property Value Total objects 18,148 LVIS categories 1,156 Renders per object 40 (multi-view) Format WebDataset (TAR shards) Number of shards 35 Approx. shard size ~5 GB each Total size ~182 GB Each object in the dataset has: A… See the full description on the dataset page:
Source: Hugging Face Hub (Alexator26/gobjaverse-lvis-dataset-v4). Metadata imported from the dataset’s Hub tags.