Installation Guide for Multiple Landmark Detection

Easy Install here

Setting up a virtual environment

The exact command may vary from system to system. On ubuntu 18.04, it is python3 -m venv multiagent

Activating the environment

source multiagent/bin/activate

Installing dependencies

pip3 install -q tensorflow tensorflow-gpu pandas opencv-python pillow gym IPython itk SimpleITK

Downloading tensorpack

wget -q https://github.com/tensorpack/tensorpack/archive/0.8.0.zip  
unzip -qq 0.8.0.zip

Fixing tensorpack

Fix a file in this version of tensorpack, setup.cfg. You could either copy it manually from the easyInstall directory or run cp ./setup.cfg tensorpack-0.8.0/

Installing tensorpack

python3 setup.py -q install
cd ../

Downloading and Installing tensorpack-medical

git clone -q https://github.com/ArjitJ/tensorpack-medical.git
cd tensorpack-medical
python3 setup.py -q install
cd ../

Fixing tensorpack-medical

We need to change a file in the install tensorpack-medical. We need to navigate to the directory where pip has saved the package.

cd ./multiagent/lib/python3.6/site-packages

Now, search for tensorpack_medical in this directory. Either there's a directory or a .egg file. If there is a directory skip this step.

unzip -qq tensorpack_medical-0.1.0-py3.6.egg

Now, copy the file from the easyInstall directory and paste it in the tensorpack_medical directory.

cd ../../../../
cp ./conv3d.py ./multiagent/lib/python3.6/site-packages/tensorpack_medical/models/

That's it!

NOTE: Tensorflow version depends on your CUDA. If this doesn't work try, pip3 install --upgrade tensorflow==1.8 tensorflow-gpu==1.8