This script resamples volumes in SimpleITK-readable formats using the CUDA-based B-Spline interpolation algorithm from this repository.
- Python 3
- SimpleITK
- CUDA (tested on 10 and 11)
- Set the
CUDA_PATHenvironment variable to point to CUDA (e.g./usr/local/cuda-11.0.3or/opt/cuda/) - Open a terminal in the git directory and run
./build.sh - Alternatively, manually run the commands in
build.sh - You should end up with the
libcudaresize.solibrary in the current directory
- Copy the
libcudaresize.sofile to/usr/libor another path onLD_LIBRARY_PATH - Copy the
cudaresize.pyfile somewhere convenient
Compiling against CUDA 11 gives a number of deprecation warnings; these are in the underlying CubicInterpolationCUDA repository (written for CUDA 3 and last updated for CUDA 5).