Caliban-2023_Schwartz_et_al

Running deepcell

In a mamba environment

These instructions are written using mamba as a package manager, but conda can be substituted for mamba.

mamba create --name deepcell python=3.8 tensorflow=2.8 cudatoolkit=11.8.0 ipykernel -c conda-forge
mamba activate deepcell
pip install -r requirements.txt

In a docker container

# Start a GPU enabled container on one GPUs
docker run --gpus '"device=0"' -it --rm \
    -p 8888:8888 \
    -v $PWD/notebooks:/notebooks \
    vanvalenlab/deepcell-tf:0.12.8-gpu

DeepCell API Key

An API key is required to access the DynamicNuclearNet dataset and Caliban models. Please see the docs for more information.

Data

The DynamicNuclearNet dataset can be accessed through deepcell.datasets (docs). Instructions for accessing additional data needed to run the notebooks and scripts in this repo are located in the README.md files inside data.

Training Caliban Models

The scripts for training the nuclear segmentation and tracking models used in Caliban are included in the training/segmentation and training/tracking folders. The deepcell environment created above can be used for all of these scripts. Instructions for running the scripts are located in training/README.md.

Benchmarking

Instructions and code for reproducing model benchmarking are included in the benchmarking folder with specific instructions for each model located in each subfolder.