BF: removed ftmp variable #534
Conversation
|
Test failure on travis :) |
| @@ -400,7 +400,7 @@ def eudx_both_directions(cnp.ndarray[double,ndim=1] seed,\ | |||
| #check for boundaries | |||
| tmp=ps[i]+step_sz*dx[i] | |||
| #ftmp=floor(tmp+.5) | |||
There was a problem hiding this comment.
Remove commented out line?
|
@arokem with the current change of this PR I get a failure from a life test. EuDX will have a slightly different output only near the boundaries of the image. Because your test is using a small ROI this difference is visible. And that is why this command fails with: |
|
Hmm. Seems best to rerun the matlab code with these new tracks. Sorry for holding this up - I will take a look tomorrow. |
Update Changelog
|
BTW - is this really for 0.8, or are we planning to do this for 0.9? We do intend to have 0.9 come down before the end of February, so it's not ridiculous to plan to solve it for that, instead of right now. |
|
This is resolving an issue reported for quite some time. Let's fix it now if possible. |
|
OK - I am working on it. On Fri, Jan 2, 2015 at 12:33 PM, Eleftherios Garyfallidis <
|
…to Garyfallidis-ftmp_remove
Used these gists: Matlab side: https://gist.github.com/arokem/5c6a28707273144404ab Python side: https://gist.github.com/arokem/8d132c35ee94cb564a73
|
OK - I am done working on this, and I will make a PR with the new files For future reference, the code to create the matlab weights and rmse is https://gist.github.com/arokem/5c6a28707273144404ab The following IPython notebook takes the resulting files, compares to dipy, https://gist.github.com/arokem/8d132c35ee94cb564a73 On Fri, Jan 2, 2015 at 12:34 PM, Ariel Rokem arokem@gmail.com wrote:
|
Garyfallidis ftmp remove
|
Can someone merge this? |
Removed unreferenced ftmp variable. This is in response to issue #449. The bug had no real effect because EuDX has already very hard constrains on the boundaries. It doesn't process seeds on the boundary voxels because of limitations with the trilinear interpolation being used. Test showing that added. Thx to the tracking team with the new tracking API we have a more flexible API to deal with the boundaries.