Skip to content

hyp-nerf/HyP-NeRF

Repository files navigation

🚀HyP-NeRF: Learning Improved NeRF Priors using a HyperNetwork

NeurIPS 2023

Bipasha Sen* 1 Gaurav Singh* 1, Aditya Agarwal* 1, Rohith Agaram 1, Madhava Krishna 1, Srinath Sridhar 2

*denotes equal contribution, 1 International Institute of Information Technology Hyderabad, 2 Brown University

applications_teaser.mp4

This is the official implementation of the paper "HyP-NeRF: Learning Improved NeRF Priors using a HyperNetwork" accepted at NeurIPS 2023

👉 TODO

  • Release pretrained checkpoints.
  • Code Release
    • Training Code
    • Architecture modules, renderer, Meta MRHE
  • ...

CREATING THE ENVIRONMENT

Please follow the steps outlined in torch-ngp repository for creating the environment upto and including the Build extension subheading.

Dataset

Download the ABO Dataset. We use the images and the transforms from abo-benchmark-material.tar and the metadata file abo-listings.tar for training. Place them in a directory structure as follows:

dataset_root
├── ABO_rendered
│   ├── B00EUL2B16
│   ├── B00IFHPVEU
│   ...
│
└── ABO_listings
  └── listings
      └── metadata
          ├── listings_0.json.gz
          ...
          └── listings_f.json.gz

Training

To train a model on the ABO dataset run the following command:

CUDA_VISIBLE_DEVICES=0 python main_nerf.py <dataset_root> --workspace <workspace dir> --bound 1.0 --scale 0.8 --dt_gamma 0 --class_choice CHAIR --load_ckpt

Testing

To render a specific NeRF from the codebook, run the following command:

CUDA_VISIBLE_DEVICES=0 python main_nerf.py <dataset_root> --workspace <workspace dir containing the pretrained ckpt> --bound 1.0 --scale 0.8 --dt_gamma 0 --class_choice CHAIR --load_ckpt --test --test_index <index of codebook>

👏 Acknowledgement

Some parts of the code are inspired and borrowed from torch-ngp (which we use as our backbone) and INR-V. We thank the authors for providing the source code.

📜 BibTeX

If you find HyP-NeRF useful in your work, consider citing us.

@article{hypnerf2023,
  title={HyP-NeRF: Learning Improved NeRF Priors using a HyperNetwork},
  author={Sen, Bipasha and Singh, Gaurav and Agarwal, Aditya and Agaram, Rohith and Krishna, K Madhava and Sridhar, Srinath},
  journal={NeurIPS},
  year={2023}
}