[Insight-developers] Change in ITK[master]: WIP: Accumulated changes to Transform and effected classes.

[Rupert Brooks]

hi Gang,

It looks like there was confusion and i was reviewing old code.  What
you have mailed and what is in master looks right to me.

The confusion arose, because

a) the first interation of this patch had it wrong (it had the same
code as currently appears in this change)
   WIP: Add new Registration/Refactoring module
http://review.source.kitware.com/#change,2096

b) Michael told me not to worry about it, so i stopped looking at it.

c) It got fixed in this change (in Patch Set 4), but not on the "WIP:
Add new Registration/Refactoring module"

d) I had kind of moved on from this patch and was looking at the
RegistrationRefactoring.  I didnt realize how out of date it was.

I hope that makes sense.  Sorry for the confusion.

Rupert

--------------------------------------------------------------
Rupert Brooks
rupert.brooks at gmail.com




On Mon, Aug 15, 2011 at 17:25,  <songgang at seas.upenn.edu> wrote:
> Hi Rupert,
>
> GetJacobian() simply calls GetJacobianWithRespectToParameters(). It uses
> GetJacobianWithRespectToPositions() in the chain rule for the composite
> transform. I put the equations in the comments. Here is the summary:
>
> * T(p1, p2, p3, x0)
> * = T3(p3, T2(p2, T1(p1, x0)))
> * dT/dp
> * = (dT/dp1, dT/dp2, dT/dp3)
> * = ( ( dT3/dT2 | x2 ) * ( dT2/dT1 | x1 ) * ( dT1/dp1 | x0 ),
> *     ( dT3/dT2 | x2 ) * ( dT2/dp2 | x1 ),
> *     ( dT3/dp3 | x2 )
>
> Could you let me know if this is the function you think not properly
> implemented?
>
> Thanks
>
> -Gang
>
>
>
>
>
>
> Mon 15 Aug 2011 04:00:09 PM EDT, Michael Stauffer <mstauff at verizon.net>
>
>
>
>  Hi,
>
> I've cc'ed Gang since this is his code.
>
> But I do see use of GetJacobianWithRespectToPosition, but I don't know
> the details.
>
> -M
>
> On 08/14/11, Matt McCormick<matt.mccormick at kitware.com> wrote:
> Hi Rupert,
>
>>
>> I think one JIRA issue may have been missed.  Based on my mental
>
> math
>>
>> (which could certainly be flawed) i was not convinced the
>
> GetJacobian
>>
>> method in the CompositeTransform was working as it should.  Mainly
>> because i dont see how it can, without calling the
>> GetJacobianWithRespectToPosition, but it doesnt call it.  In any
>
> case,
>>
>> Michael asked me to treat this as work in progress and not do a full
>> numerical test yet.  It should be done someday though, and i am
>> notoriously forgetful.   I'm not sure how to create a JIRA issue or
>> even if i can, but we should be sure not to forget this one.
>>
>
> Thanks for the note.  I created a ticket here:
>
> https://issues.itk.org/jira/browse/ITK-2606
>
>> The registration refactoring commit
>> http://review.source.kitware.com/2096 is also a huge patch set that
>> changes a lot of stuff.  It risks to suffer from the same problem.
>
> If
>>
>> theres any way at all to split it up into pieces that might make the
>> reviews go faster.  Ideally, I would like to be sure i understand
>
> the
>>
>> design intent first - then i can check that the code actually
>
> delivers
>>
>> what the design intends that it should.
>>
>> I feel i should point out that what makes these so tough to review
>
> is
>>
>> there is a massive amount of very good, detailed work involved.   To
>> give it the thorough review it deserves is inevitably time
>
> consuming,
>>
>> I hope those actually doing the work arent getting discouraged by
>
> all
>>
>> the questioning.
>>
>
> I concur.  It is also good to start getting the code merged so
> dashboard issues and interactions with the rest of the toolkit can be
> addressed.  Thanks to Luis, Bill L, and Brian for the help this
> weekend fixing up the dashboard.
>
> Matt
> _______________________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Kitware offers ITK Training Courses, for more information visit:
> http://kitware.com/products/protraining.html
>
> Please keep messages on-topic and check the ITK FAQ at:
> http://www.itk.org/Wiki/ITK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.itk.org/mailman/listinfo/insight-developers
>
>
>