3D / Point Cloud
Objaverse
800K+ annotated 3D models for training 3D and multimodal models.
100K–1M·CC-BY·Point Cloud (LAS/PLY)
3D / Point Cloud
libero-3d-pointcloud
LIBERO 3D Point-Cloud GT (sim teacher) Per-frame, object-level 3D point-cloud ground truth for the LIBERO benchmark, generated by replaying the…
LIBERO 3D Point-Cloud GT (sim teacher) Per-frame, object-level 3D point-cloud ground truth for the LIBERO benchmark, generated by replaying the official LIBERO human demonstrations’ recorded MuJoCo states in simulation and unprojecting the metric depth + per-object GT instance segmentation (the “MuJoCo privileged upper bound”). No estimation — exact metric geometry. Source: LIBERO base demos (yifengzhu-hf/LIBERO-datasets), 130 tasks × 50 demos across the 5 suites (spatial /… See the full description on the dataset page:
Source: Hugging Face Hub (Kit-Key/libero-3d-pointcloud). Metadata imported from the dataset’s Hub tags.