Code for Closing the Visual Sim-to-Real Gap with Object-Composable NeRFs

This repo contains code pertaining to our ICRA 2024 paper (arxiv):

"Closing the Visual Sim-to-Real Gap with Object-Composable NeRFs".
Nikhil Mishra, Maximilan Sieb, Pieter Abbeel, Xi Chen.
In the proceedings of the IEEE International Conference on Robotics and Automation (ICRA), 2024.

The object-centric NeRF model we proposed, COV-NeRF, is implemented in cov_nerf/model.py.

The Jupyter notebook example.ipynb shows how to render and manipulate scenes with a pretrained COV-NeRF model we have provided.

Run download.py to fetch the checkpoint and data from Google Drive.

BibTeX

 @inproceedings{
     mishra2024closing,
     title={Closing the Visual Sim-to-Real Gap with Object-Composable NeRFs},
     author={Nikhil Mishra and  Maximilian Sieb and Pieter Abbeel and Xi Chen},
     year={2024},
     booktitle={IEEE International Conference on Robotics and Automation (ICRA)},
}

License

Shield:

This work, including the paper, code, weights, and dataset, is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.