This Repository contains a development framework for using tensorflow for labeling 3D neuroimaging voxel data. It was developed during BrainHack Boston 2017. This is a work in progess. It was based off a 2D pixel labelling example given by Nvidia at a Harvard Compute fest in January 2017. The referece code is in directory ./ref-example
Setup:
git clone https://github.com/pwighton/neuroimage-tensorflow
cd neuroimage-tensorflow
curl -o b40.tar.gz 'https://gate.nmr.mgh.harvard.edu/filedrop2/index.php?p=1m8hsmv9nkj'
tar zxvf b40.tar.gz
cd ./bucker40/
mkdir train
mv 004 ./train
mv 008 ./train
cd..
Build the docker container
docker build --no-cache -t tensorflow-tensorboard-nibabel ./docker/
Start the docker container (with tensorflow, tensorboard and jupyter), mapping ports for tensorboard and jupyter and mounting the repo dir into /notebooks/data
docker run -it --rm -p 8888:8888 -p 6006:6006 -v ${PWD}:/notebooks/data tensorflow-tensorboard-nibabel
The jupyter URL is shown in the docker terminal window and should look something like
http://localhost:8888/?token=bca7c05e6447dc94a80e895bb8e97eb811218e6427af8c12
The tensorboard URL is
http://localhost:6006/
You should now be able to step through the neuro-example.ipynb
notebook