This is a repository to run EVAC+: Multi-scale V-net with Deep Feature CRF Layers for Brain Extraction
This model depends on the implementation of the CRFasRNNLayer, which is from this repository : https://github.com/MiguelMonteiro/CRFasRNNLayer.
Files were copied for convenience.
Please follow the instructions of that repo which is also written below.
===================================================
Add the correct paths to nvcc, g++ and CUDA/Include in build.sh file found inside permutohedral_lattice folder. (lines 8, 9, and 10).
Example :
CUDA_COMPILER=/usr/local/cuda/bin/nvcc
CXX_COMPILER=/usr/bin/g++
CUDA_INCLUDE=/usr/local/cuda/include/
Create a symlink using the following code so that libtensorflow_framework.so points to libtensorflow_framework.so.2:
ln -s path_to_libtensorflow_framework.so.2 path_to_libtensorflow_framework.so
Then compile the code using:
sh build.sh
Running the above command in the main folder gives you lattice_filter.so file. Please make sure that this file is present. If not, the models.py code would throw errors.
See the nested module permutohedral_lattice for more information on compilation for different image types.
====================================================
Just running the command should be enough to run the model.
We have currently tested the code only on Ubuntu.
Please refer to training.ipynb and testing.ipynb.
Specifications of the model can be found in models.py