Deep learning based Image Compression for Microscopy Images

Introduction

This repo is the code base for the paper: Deep learning based Image Compression for Microscopy Images: An Empirical Study

The submodule CompressAI is forked from CompressAI and adapted to the grayscale and 2/3D microscopy images.

Installation

• clone the repo:

  git clone --recurse-submodules https://github.com/MMV-Lab/data-compression.git
  cd CompressAI
  

• install the environment (requires conda):

  conda create -n CompressAI python=3.9
  conda activate CompressAI
  pip install -e .    
  

• also need to install the mmv_im2im package if you want to do the downstream labelfree task. Please check the github repo for the installation.

Reproducibility

You can go to this folder and try out our jupyter notebooks for both 2D and 3D tasks.

We will release the pretrained models and dataset to Zenodo.

Acknowledgement

This project is the application and adaptation of the CompressAI tool in the bioimage field.

Citation

@article{zhou2023deep,
  title={Deep learning based Image Compression for Microscopy Images: An Empirical Study},
  author={Zhou, Yu and Sollmann, Jan and Chen, Jianxu},
  journal={arXiv preprint arXiv:2311.01352},
  year={2023}
}