Speed up the reconstruction

Yes , I hve tried this method, and run .\windows-download-cache-and-build-module-wheels.ps1 -python_version_minor 9 -setup_options "--exclude-libs nvcuda.dll" -cmake-options "-DRTK_USE_CUDA:BOOL=ON"


But I don’t konw where to set (ITK_WRAP_IMAGE_DIMS:STRING=2;3)? And on the .ps1 file the verion can only set as { $env:ITK_PACKAGE_VERSION = 'v5.4.0' } not { $env:ITK_PACKAGE_VERSION = 'v5.4rc1' }

I’m assuming you download the ITKPythonPackage repository locally so there is nothing that prevents you from modifying ITK_WRAP_IMAGE_DIMS option and ITK_PACKAGE_VERSION in the scripts.

I use git clone to pull ITKPythonPackage. When I search the methods that meet my needs, there are different ways, such as use windows-download-cache-and-build-module-wheels.ps1 or setup.py or windows_build_wheels.py. But they all failed. :sob:
Now I use windows_build_wheels.py to generate wheels because I can set some parameters on the code. But when I run this, the error occurs as below:

I tried to find out what caused it, but there was no progress. :sob:

Good News, I succeed generate itk wheels! The next step is to generate itk-rtk-cuda(I have tried use the wheels to pip itk and use pip install it-rtk-cuda, but there has the error beacause the image dims).
The tips is that you need make sure the compilation environment is consistent with your requirements, and my computer defaults to x86 but I need x64.

@zhangm thanks for sharing your updates and success!

Just a note that @simon.rit recently released itk-rtk 2.6.0 binaries based on itk-5.4.0.

1 Like