Revisiting Implicit Neural Representations in Low-Level Vision

Paper | Project Page

This repository provides the implementation of the paper:

Wentian Xu and Jianbo Jiao. Revisiting Implicit Neural Representations in Low-Level Vision. In ICLR-NF (2023).

We provide Jupyter Notebooks for each task in the paper. We also provide a denosing example based on .py file.

Setup

To install the dependencies into a new conda environment, simply run:

conda env create -f environment.yml
source activate linr

Alternatively, you may install them manually:

conda create --name linr
source activate linr
conda install python=3.9
conda install pytorch=1.12 torchvision cudatoolkit=11.6 -c pytorch
conda install numpy
conda install matplotlib
conda install scikit-image
conda install jupyter

How to run

Change the file path in jupyter Notebooks or in train_denoising.py to the image you want
Run each cell in jupyter Notebooks or run the code in train_denoising.py

Data

The data used in the paper is stored in the Data folder. You can also use your own data by changing the file path.

Citation

If you find LINR useful, please cite the following BibTeX entry:

@inproceedings{linr,
  title={Revisiting Implicit Neural Representations in Low-Level Vision},
  author={Wentian Xu and Jianbo Jiao},
  booktitle="International Conference on Learning Representations Workshop",
  year={2023},
}

Achnowledgement

Part of the codes is based on DIP, SIREN and DeepRED.

WenTXuL / LINR