[Nipy-devel] data structures for brain surfaces

[Fernando Perez]

On 3/13/07, Darren Weber <dweber@radmail.ucsf.edu> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
>
> Thanks, Michael.
>
> What is your view of how vtk handles data streams?  I'm concerned that
> copies of large data arrays through each stage of processing will exceed
> memory limits.  Do you handle most of the data in python lists or numpy
> and only convert into vtk for display?

You should have a look at TVTK, which has really clean numpy
integration without copies (so you can do numerics on the actual
memory zone VTK is displaying).  Besides, it's a LOT nicer to code
against than the regular (Kitware-supplied) python VTK API.

I recently started actually using it (after toying with it for a
while) and I'm quite impressed.

Cheers,

f