This project tests four major segmentation networks on the BSData dataset. To setup the project for execution: Download the dataset in question
git clone https://github.com/2Obe/BSData.git
Install requirements
pip install -r requirements.txt
create necessary folders
mkdir IMAGES MASKS IMAGE_PATCHES Masks_PATCHES IMAGE_PATCHES_AUGUMENTED
Convert json to png masks
python json2mask.py
Transform images into patches
python patching_and_saving.py
Train the network on default parameters, signing into Wandb will be necessary for progress logging
python Unet.py
To train the model without wandb integration run
python UNet-noWandb.py