PNL_TRAIN.md

CNN-Diffusion-MRIBrain-Segmentation repository is developed by Senthil Palanivelu, Suheyla Cetin Karayumak, Tashrif Billah, Sylvain Bouix, and Yogesh Rathi, Brigham and Women's Hospital (Harvard Medical School).

Training CNN at PNL

Training the CNN on new data requires a reasonably powerful GPU machine. Two such machines are available at PNL: pnl-oracle and pnl-maxwell. You can ssh into them as follows:

ssh pnl-oracle
ssh pnl-maxwell

You may need to append .bwh.harvard.edu suffix at the end of the above hostnames.

Then, source the following environment that makes use of tensorflow-gpu:

source /rfanfs/pnl-zorro/software/pnlpipe3/CNN-Diffusion-MRIBrain-Segmentation/train_env.sh

Finally, follow the instruction from README.md#training to perform training.