license: odc-by
language:
- en
viewer: false
tags:
- huggingscience
- science
- computer-vision
- 3d
- robotics
- embodied-ai
- graphics
datasets:
- allenai/objaverse
Objaverse
Overview
Welcome to Objaverse, the world's largest 3D object dataset. With over 800,000 high-quality annotated 3D objects, this groundbreaking collection is designed to advance research in 3D vision, robotics, and artificial intelligence.
Dataset Scale
Setting a new standard in 3D data resources, Objaverse is 100x larger than all other 3D datasets combined, enabling unprecedented opportunities for research and development in 3D understanding and generation.
Dataset Description
Dataset Summary
Objaverse is a large-scale dataset of 3D models aimed at filling a critical gap in AI research by providing diverse, high-fidelity 3D data. The dataset includes:
- 800,000+ 3D models
- Descriptive captions and tags for each object
- Animations for dynamic objects
- Multiple object categories with rich intra-category diversity
- High-quality renderings and metadata
Supported Tasks
The dataset supports various research applications including:
- Training generative 3D models
- Improving tail category segmentation (LVIS benchmark)
- Training open-vocabulary object-navigation models for Embodied AI
- Creating benchmarks for robustness analysis of vision models
- Supporting research in computer graphics and 3D vision
Dataset Size and Format
- Total Objects: 800,000+
- Format: GLB/GLTF files
- Additional Metadata: JSON files with annotations and licenses
Data Collection and Curation
The dataset has been curated to ensure:
- High-quality 3D models
- Diverse object categories
- Rich metadata and annotations
- Proper licensing and attribution
Dataset Usage and Getting Started
Quick Start Resources
- 📝 Paper - Technical details and methodology
- 🌐 Website - Interactive demos and visualizations
- 🤗 Hugging Face Dataset - Download and use the dataset
- 📓 Google Colab Tutorial - Interactive tutorial to get started
- 💻 GitHub Repository - Code and scripts
Integration Examples
Each Objaverse object can be rendered and imported into Blender. See the rendering scripts to get started.
Model Applications
The dataset has enabled training of powerful 3D foundation models like Zero123-XL, demonstrating impressive 3D generation capabilities. Key applications include:
- 🎨 3D Generation: Train and use generative models for 3D content creation
- 🔄 Single Image to 3D: Convert single images to 3D models using Zero123-XL
- 🤖 Robotics: Train embodied AI agents for object interaction and navigation
- 🖼️ Computer Vision: Improve object detection and segmentation tasks
License
The use of the dataset as a whole is licensed under the ODC-By v1.0 license. Individual objects in Objaverse are all licensed as creative commons distributable objects, and may be under the following licenses:
- CC-BY 4.0 - 721K objects
- CC-BY-NC 4.0 - 25K objects
- CC-BY-NC-SA 4.0 - 52K objects
- CC-BY-SA 4.0 - 16K objects
- CC0 1.0 - 3.5K objects
The metadata will provide the license for each object.
Citation
To cite Objaverse, please use the following BibTeX entry:
@article{objaverse,
title={Objaverse: A Universe of Annotated 3D Objects},
author={Matt Deitke and Dustin Schwenk and Jordi Salvador and Luca Weihs and
Oscar Michel and Eli VanderBilt and Ludwig Schmidt and
Kiana Ehsani and Aniruddha Kembhavi and Ali Farhadi},
journal={arXiv preprint arXiv:2212.08051},
year={2022}
}