/DeepCell-Keras

Reimplement Deep Cell with Keras and Horovod.

Primary LanguagePythonMIT LicenseMIT

Deep Cell

Reimplement Deep Cell with Keras and Horovod. The paper of Deep Cell is here.

Dataset

Download the Dataset:

./get_dataset.sh

Enviroment

If you use pip to manage your Python enviroment:

pip install -r requirements.txt

If you use conda as your Python enviroment:

conda env create -f environment.yml

Usage

The sub-command can be

  1. data_prepare
  2. train
  3. validation
  4. test

Or you can use pipeline as sub-command to run all 4 steps.

python launch.py pipeline

Tips

  • We use "channels_first" date format in Keras, you should config in ~/.keras/keras.json.

  • If you want to run a distributed training, you must install Horovod and run

     mpirun -np 4 python deepcell/train.py -e 20 --dist 1

Thanks

DeepCell and deepcell-tf