Skip to content

[Code Optimisation] Adding GPU Support to existing Random Walker with Prior Implementation

Notifications You must be signed in to change notification settings

aymuos15/GPU_Random_Walker_with_Prior

Repository files navigation

GPU Implementation of Random Walks with Prior

helpers: Contains the scripts to run everything. helper_numpy is almost an identical copy paste of the original repo.

visualisation.ipynb: Contains a notebook giving visual unit test.

Performance Results:

GPU Speed up for lapcian @size: (125, 125, 125) 63.231481243642904 % (just run python test.py)

Notes:

  • Have to create more unit tests in terms of visualisation

To-do:

  • Create a GPU based approach with torch instead of numpy on CPU
  • Unit Tests
  • Input and Output Testing with mock nifti like files.
  • VERY POOR Memory optimisation at the moment.

References :

From the vs runs of aaron
  • vsmcrc_102608_0000.nii.gz* -> Raw Volume
  • vsmcrc_102608.nii.gz -> Segmentation result from model-1 910-910
  • vsmcrc_102608.npz -> Weights for prior from model-1 910-910

About

[Code Optimisation] Adding GPU Support to existing Random Walker with Prior Implementation

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published