To reproduce the results run
python train_scnym --dataset {dataset} --config {config} --basedir {path to save models}
The dataset in the command above can be rat for the Rat dataset and hb for the hPMBC dataset.
The config in the above command controls the models as described below:
- no_ssl: Source Only
- only_dan: DANN
- no_dan: MixMatch
- no_new_identity: MixMatch+DANN
The plots can be generated using the scnym_atlas_transfer.ipynb notebook.
To reproduce results run
python coarse_trainer.py
python nocoarse_trainer.py
To produce plots run the following Jupyter notebook
Notebooks/scnym_atlas_transfer-hpbmc-run.ipynb
to reproduce results, run
python bioTrain.py --gpu_ids=[*gpu -ids here*]
Then go through scnym_atlas_transfer_cycleGAN.ipynb and run cells