inrlib

Zach Stoebner

This library is intended for anyone seeking to train an implicit neural representation (INR) on discretized signal measurements. It contains basic implementations of a neural implicit MLP, losses, constraints, and logging for complex-valued data. Refer to the demo to get started.

Setup

Create a dedicated conda environment called inr:

conda env create -f environment.yml

This environment assumes Linux and CUDA>=12.1.

If the file doesn't work out of the box, the key requirements are:

• python>=3.10
• pytorch
• lightning

For the demo, additionally run:

pip install gdown phantominator omegaconf

Usage

sh run_from_config.sh <STAGE=[train, val, test, pred]> <CONFIG=path/to/config> <GPUID=int>

Refer to the demo for example implementation and how to extend to new models, losses, constraints, etc. Refer to the complex demo for an example with a complex-valued neural network.

References

@article{sitzmann2020implicit,
  title={Implicit neural representations with periodic activation functions},
  author={Sitzmann, Vincent and Martel, Julien and Bergman, Alexander and Lindell, David and Wetzstein, Gordon},
  journal={Advances in neural information processing systems},
  volume={33},
  pages={7462--7473},
  year={2020}
}

@article{tancik2020fourier,
  title={Fourier features let networks learn high frequency functions in low dimensional domains},
  author={Tancik, Matthew and Srinivasan, Pratul and Mildenhall, Ben and Fridovich-Keil, Sara and Raghavan, Nithin and Singhal, Utkarsh and Ramamoorthi, Ravi and Barron, Jonathan and Ng, Ren},
  journal={Advances in Neural Information Processing Systems},
  volume={33},
  pages={7537--7547},
  year={2020}
}

@article{mildenhall2021nerf,
  title={Nerf: Representing scenes as neural radiance fields for view synthesis},
  author={Mildenhall, Ben and Srinivasan, Pratul P and Tancik, Matthew and Barron, Jonathan T and Ramamoorthi, Ravi and Ng, Ren},
  journal={Communications of the ACM},
  volume={65},
  number={1},
  pages={99--106},
  year={2021},
  publisher={ACM New York, NY, USA}
}