GLIMPSE: Generalized Local Imaging with MLPs

This repository is the official Pytorch implementation of "GLIMPSE: Generalized Local Imaging with MLPs".

Colab demo

Requirements

(This code is tested with PyTorch 1.12.1, Python 3.8.3, CUDA 11.6 and cuDNN 7.)

• numpy
• scipy
• matplotlib
• imageio
• torch==1.12.1
• torchvision=0.13.1

Installation

Run the following code to install conda environment "environment.yml":

conda env create -f environment.yml

Dataset

All datasets have been uploaded to SwitchDrive. You can access the complete LoDoPaB-CT by downloading it from here. Additionally, we have made available a smaller subset of the LoDoPaB-CT dataset, comprising approximately 1000 training and 100 test samples. Moreover, to evaluate model generalization, we have included out-of-distribution (OOD) brain images consisting of 18 samples. These datasets can be downloaded using the following commands:

Complete LoDoPaB-CT:

curl -O -J https://drive.switch.ch/index.php/s/XzMbtHQFrQsLgxC/download

Small LoDoPaB-CT training subset:

curl -O -J https://drive.switch.ch/index.php/s/qMlALcE7AZzUPBh/download

Small LoDoPaB-CT test subset:

curl -O -J https://drive.switch.ch/index.php/s/fWBUmtZjozwpN9W/download

Out-of-didstribution brain images:

curl -O -J https://drive.switch.ch/index.php/s/BQ8Yb8ofjutsEjV/download

After downloading the datasets, please sepcify the training, test and OOD directories in 'config.py' script.

Experiments

Training & Inference

All arguments for training are explained in 'config.py'. After specifying your arguments, you can run the following command to train the model:

python3 train.py 

Citation

If you find the code useful in your research, please consider citing the paper.

@article{khorashadizadeh2024glimpse,
  title={GLIMPSE: Generalized Local Imaging with MLPs},
  author={Khorashadizadeh, AmirEhsan and Debarnot, Valentin and Liu, Tianlin and Dokmani{\'c}, Ivan},
  journal={arXiv preprint arXiv:2401.00816},
  year={2024}
}