3D / Point Cloud
Objaverse
800K+ annotated 3D models for training 3D and multimodal models.
100K–1M·CC-BY·Point Cloud (LAS/PLY)
3DEdit-1M A large-scale 3D edit-pair dataset of the paper Omni123: Exploring 3D Native Foundation Models with Limited 3D Data by Unifying Text to 2D and 3D Generation. Each sample is a paired (source, target) 3D edit. The pipeline behind every sample: Edit instruction synthesized by Qwen3.5-35B-A3B. Target image (post-edit RGB render) generated by FLUX.2-klein-9b-kv. Source and target 3D meshes generated by Hunyuan3D-2-mini from the source and target images. Shape tokenization: each mesh is… See the full description on the dataset page:
Source: Hugging Face Hub (OmniWeave/3DEdit-1M). Metadata imported from the dataset’s Hub tags.