So the tract features which are given in the training dataset are extracted from the code by the concept discussed in the paper DeepWMA, feature_RAS and the flipping and copying of the streamlines.
The preprocessing script seems to be the same for converting the nii.gz -> nrrd -> vtk -> h5
So the tract features which are given in the training dataset are extracted from the code by the concept discussed in the paper DeepWMA, feature_RAS and the flipping and copying of the streamlines. The preprocessing script seems to be the same for converting the nii.gz -> nrrd -> vtk -> h5