ENH: added benchmark for ITK ResampleImageFilter #56
Conversation
The benchmark includes support for 1D/2D/3D images, common interpolators (Linear, Nearest, BSpline), common registration transforms and their combinations (Affine, BSpline, Euler, Similarity, Translation, Identity) and other options to fully benchmark Resample for registration tasks.
| // | ||
| // Command line argument "-i" "Linear" ["Nearest"] ["BSpline"] | ||
| // Command line argument "-soi" controls spline order interpolator for the BSpline interpolator. | ||
| // For "-i" "BSpline" the 0th - 5th order splines are supported, the default 3th. |
There was a problem hiding this comment.
Thanks, I will fix the typos.
| } | ||
|
|
||
| //------------------------------------------------------------------------------ | ||
| // This helper function completely define the transform[s] |
| } | ||
|
|
||
| //------------------------------------------------------------------------------ | ||
| // This helper function completely set the single transform |
There was a problem hiding this comment.
This is awesome @dpshamonin !
As per the ITK style (section C of the Appendix in [1]), I'd rather:
- Align the loop/conditional statements to the curly brackets.
- Try to remove the white spaces before the loop/conditional statements and the brackets.
- Remove the unnecessary
//---------------------------------------
comment lines.
- In order to embrace C++11, use range-based loops.
- Although, this has also been on my ToDo list for a long, long time, many of the main program's argument processing/checking functions could be re-used by other testing/benchmarking code, so I guess one day or another it would be useful to create a utils class to gather all this.
Thanks.
|
@jhlegarreta |
|
Great! ✨ |
|
Thanks Dennis. |
|
🥇 Thanks, @dpshamonin ! |
|
Great! this is awesome, but now the filtering label takes quite a bit of time. Should we create a separate label and CMake option for for Resample? Opened #59 |
The benchmark includes support for 1D/2D/3D images,
common interpolators (Linear, Nearest, BSpline),
common registration transforms and their combinations
(Affine, BSpline, Euler, Similarity, Translation, Identity)
and other options to fully benchmark Resample for registration tasks.