This repository contains two module including residual contextual and subpixel convolution network (RC-SPCNet) and post-processing (lifted multicut), developed by Hainan university and Institute of Automation, Chinese Academy of Sciences.
Environment
The code is developed and tested under the following configurations.
• Network
CUDA>=11.0, Python>=3.7, tensorflow-gpu>=1.14.0, keras>=2.2.5
• Lifted multicut
Python>=3.7, nifty, elf, vigra, numpy, imageio, napari
Installation
We recommend creating new environments independently.
• Download the repository
• Install the required python environment
• Activate your new environment and run the code
For the post-processing method, we suggest install the package via conda and pip as follow:
conda install -c conda-forge vigra
conda install -c conda-forge nifty
conda install -c conda-forge python-elf
python -m pip install imageio
python -m pip install "napari[all]"
Test Data
The dataset of lifted multicut can be download in： https://pan.baidu.com/s/1_fX2FNKDXMMSWe8pQOnh4g, code：ydlo
How to use (lifted multicut)
Adjust the data path and run the code “postprocess_lifted_multicut.py”, the result will be saved in the specified path.
Acknowledgement
This project is built upon some previous projects. Especially, we'd like to thank the contributors of the following github repositories: multicut_pipeline: https://github.com/ilastik/nature_methods_multicut_pipeline and elf: https://github.com/constantinpape/elf
Contact
If you meet any problem, please contact the author directly.
Email: [email protected]