[Rtk-users] Reconstruction filter progress observer

[MURIENNE Albert]

Cyril, Simon,

Thanks for your detailed answers. I’ll try to give a go with the FDK filter first, and I’ll keep you updated.

Regards,

Albert

De : Cyril Mory [mailto:cyril.mory at creatis.insa-lyon.fr]
Envoyé : jeudi 9 novembre 2017 14:34
À : MURIENNE Albert
Cc : rtk-users at public.kitware.com
Objet : Re: [Rtk-users] Reconstruction filter progress observer


A few more sentences on this point:

- Indeed, very few RTK filters currently perform progress reporting

- For pipelines (instead of individual filters), you want to use that class: https://itk.org/Doxygen/html/classitk_1_1ProgressAccumulator.html.

- Despite its name, "progress.CompletedPixel()" can be called after any meaningful step, not just the update of a pixel. On iterative filters, for example, http://www.cs.cmu.edu/~galeotti/methods_course/ITK_Filters2.pdf suggest that ProgressReporter can be initialized with the maximum number of iterations instead of with the total number of pixels to update, and progress.CompletedPixel() can be called at each iteration. So if you are reconstructing with an iterative reconstruction filter, you could start by trying to add a ProgressReporter to that class, and count its iterations. That's for the modifications to perform in the RTK code, it is probably just a few lines for the first test.

Now, your question was about the existence of some code to observe the progress of a filter, assuming that filter does call the
progress.CompletedPixel() adequately. You may want to check this ITK example:
https://itk.org/Doxygen/html/SphinxExamples_2src_2Core_2Common_2ObserveAnEvent_2Code_8cxx-example.html

I hope it helps,
Cyril

On 09/11/2017 14:29, Simon Rit wrote:
Dear Albert,
We have not implemented anything in this direction so far. But it would be useful to have it and I know there is a mechanism to do that in ITK Have you ever used it on your side?
I think the idea is:
- in the reconstruction filter, to create and regularly call a ProgressReporter<https://itk.org/Doxygen/html/classitk_1_1ProgressReporter.html>,
- where you want to observe the progress, you add an observer.
There are many examples in ITK. If you want to do it, feel free to give it a try. If you need help, I'm happy to support your. I have created an issue<https://github.com/SimonRit/RTK/issues/114> in github to keep it in mind.
Simon

On Thu, Nov 9, 2017 at 1:27 PM, MURIENNE Albert <albert.murienne at thalesgroup.com<mailto:albert.murienne at thalesgroup.com>> wrote:
Cyril, to be more precise : if I run a long time reconstruction, I want to be available to track progression of the computation (ie 0 to 1 floating index), to have an idea of how fast it is running, if we are close to the end and so on…

Albert

De : Cyril Mory [mailto:cyril.mory at creatis.insa-lyon.fr<mailto:cyril.mory at creatis.insa-lyon.fr>]
Envoyé : jeudi 9 novembre 2017 13:25
À : MURIENNE Albert; rtk-users at public.kitware.com<mailto:rtk-users at public.kitware.com>
Objet : Re: [Rtk-users] Reconstruction filter progress observer


Hello,

I am not sure I understand. Are you trying to know, during runtime, which filter is currently being updated ?

Otherwise, can you reformulate your question (possibly with more context) ?
Best,
Cyril
On 09/11/2017 13:11, MURIENNE Albert wrote:
Hello,

Is there a simple code example available to illustrate the registration of an observer to be notified with the progression of rtk’s reconstruction filters?

Thanks in advance for your help,

Regards

--

Albert Murienne



_______________________________________________

Rtk-users mailing list

Rtk-users at public.kitware.com<mailto:Rtk-users at public.kitware.com>

http://public.kitware.com/mailman/listinfo/rtk-users


_______________________________________________
Rtk-users mailing list
Rtk-users at public.kitware.com<mailto:Rtk-users at public.kitware.com>
http://public.kitware.com/mailman/listinfo/rtk-users


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/rtk-users/attachments/20171110/dd9238ea/attachment.html>