facebookresearch/uco3d resource

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1343 stars · 186 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

uCO3D (UnCommon Objects in 3D) is a large-scale dataset from Meta AI containing ~170,000 turn-table videos of objects from ~1000 LVIS categories, annotated with segmentation, camera poses, point clouds, 3D Gaussian Splats, and captions. This repository provides download scripts and tooling for working with the dataset, which is hosted on Hugging Face.

Use cases

• download a large-scale 3D object video dataset
• train 3D reconstruction or novel view synthesis models
• get camera poses and point clouds for object videos
• obtain 3D Gaussian Splat reconstructions of real objects
• find multi-category 3D object data with captions
• preview a small subset before committing to a 19 TB download

When to choose

• you need large-scale real-world 3D object data for training generative or reconstruction models
• you want videos with rich annotations like segmentation, poses, and Gaussian Splats
• you need diverse object categories from the LVIS taxonomy with text captions

When to avoid

• you lack ~19.3 TB of storage for the full dataset
• you only need synthetic 3D data or meshes rather than real captured videos
• you need frame-based data like CO3Dv2 instead of full videos

Facets

dataset · maturity stable

data-science machine-learning computer-vision image-processing video-processing computer-vision machine-learning artificial-intelligence python windows 3d-dataset 3d-reconstruction gaussian-splatting camera-pose point-clouds lvis video-dataset facebook-research cc-by-4.0 datasets linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="facebookresearch/uco3d")