Cellulus-track

Introduction

Cellulus-track is a deep learning based method which can be used to obtain instance-segmentation and tracking of objects in 2D or 3D microscopy images in an unsupervised fashion i.e. requiring no ground truth labels during training.

Installation

One could execute these lines of code below to create a new environment and install dependencies.

1. Create a new environment called cellulus-track:

conda create -y -n cellulus-track python==3.9

2. Activate the newly-created environment:

conda activate cellulus-track

3a. If using a GPU, install pytorch cuda dependencies: I recommend using the mamba solver for installing pytorch cuda dependencies, to reduce the likelyhood of a very slow (24hr+) dependency resolve:

conda install conda-libmamba-solver
conda config --set solver libmamba

then install dependencies: (better on linux:)

conda install pytorch==2.0.1 torchvision==0.15.2 pytorch-cuda=11.7 -c pytorch -c nvidia

(better on windows:)

pip3 install torch torchvision --index-url https://download.pytorch.org/whl/cu126

3b. otherwise (if using a CPU or MPS), run:

pip install torch torchvision

4. Install the package from github:

pip install git+https://github.com/h-westmacott/cellulus-track

5. If that isn't working, for example it's throwing a undefined symbol: iJIT_IsProfilingActive error:

pip install torch==2.0.1+cu117 --extra-index-url https://download.pytorch.org/whl/cu117