This repository contains MuJoCo MJCF models for all 1030 household objects in Google's Scanned Objects Dataset. The XML files were automatically generated using obj2mjcf. A sample of the models is shown below:

In each object directory, you'll find the following:

model.obj
model_collision_*.obj  # 32 collision submeshes.
texture.png
model.xml

The model definition lives in model.xml. To visualize it, you can drag-and-drop it in the simulate binary. You can examine the visual and collision meshes by cycling between groups 2 and 3. The collision mesh is a convex approximation that has been generated with V-HACD.

If there are any problems loading a particular model in MuJoCo, please raise an issue on the GitHub repository.

The 3D assets (obj and png) are licensed under the CC-BY 4.0 License. The MJCF (XML) files are licensed under the MIT License.

If you use these models in your research, consider citing the original paper:

@misc{downs2022scannedobjects,
    url = {https://arxiv.org/abs/2204.11918},
    author = {Downs, Laura and Francis, Anthony and Koenig, Nate and Kinman, Brandon and Hickman, Ryan and Reymann, Krista and McHugh, Thomas B. and Vanhoucke, Vincent},
    keywords = {Robotics (cs.RO), Graphics (cs.GR), FOS: Computer and information sciences, FOS: Computer and information sciences},
    title = {Google Scanned Objects: A High-Quality Dataset of 3D Scanned Household Items},
    publisher = {arXiv},
    year = {2022},
}

If you're feeling generous, you can also cite this repository:

@software{zakka2022scannedobjectsmujoco,
    author = {Zakka, Kevin},
    month = {7},
    title = {{Scanned Objects MuJoCo Models}},
    url = {https://github.com/kevinzakka/mujoco_scanned_objects},
    version = {0.0.1},
    year = {2022}
}

Disclaimer: This is not affiliated with Scanned Objects or Google Research.