MitoNet: a scalable framework for automated mitochondria segmentation

Install Pytorch Connectomics

To prepare your Windows computer for the installation, first install these applications:

• Install Visual Studio
• Install CUDA Toolkit
• Install Anaconda
• Install Git

  conda install -c anaconda git
  

Now follow the steps below to install the package:

1) Open your Anaconda Prompt (click Start, then search, or select Anaconda Prompt from the menu) and run the follow commands in sequence:

conda create -n py3_torch python=3.8
activate py3_torch
conda install pytorch torchvision cudatoolkit=11.0 -c pytorch

2) Check your Pytorch version (we want the version to be >1.80):

pip3 show torch

3) Check if Pytorch is installed with CUDA support:

activate py3_torch
python

import torch
torch.cuda.is_available() # output would be True or False

• Note: if you get a False, try to reinstall pytorch with CUDA in different ways; also, you should go to the official website for detailed instructions; solutions that worked for us:

• using conda:

  conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch

• using pip:

  pip3 install torch==1.10.1+cu113 torchvision==0.11.2+cu113 torchaudio===0.10.1+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html

4) Verify if nvcc is accessible from terminal:

nvcc --version

5) Finally, install the Pytorch Connectomics package:

git clone https://github.com/zudi-lin/pytorch_connectomics.git
cd pytorch_connectomics
pip install --editable .

For installation on Linux machines, follow the instructions here.

Semantic Segmentation

Before running the model, first install wget for Windows - a free network utility to retrieve files from the World Wide Web using HTTP and FTP. This post provides a detailed tutorial for installing wget. In general, follow the steps below in sequence:

• Download wget: https://eternallybored.org/misc/wget/
• Copy wget.exe to C:\Windows\System32 folder
• In your command line, type:

  wget -h
  

Now start the training process:

1) Download the sample dataset:

wget http://rhoana.rc.fas.harvard.edu/dataset/lucchi.zip

Note: wget downloads files in the current working directory where it is run

2) Configure the model for training:

# start anaconda prompt
conda activate py3_torch
python

import torch

print(f"Is CUDA supported by this system? {torch.cuda.is_available()}") 
# Returns True if CUDA is supported by your system, else False

print(f"ID of current CUDA device: {torch.cuda.current_device()}")
# Returns ID of current device (you need this info for configuration)

3) Run the script for training:

Configure the model

Instance Segmentation