Re: [Teem-users] tensor interpolation question 2

[Nicholas W. <nw...@uc...>]

I get a similar problem of NaN's with the log euclidean code.  I've been
trying to isolate the problem but haven't yet.

Nicholas Wisniewski
University of California Los Angeles
Department of Physics and Astronomy
Department of Medicine


On Mon, Nov 23, 2009 at 5:08 AM, Gordon L. Kindlmann <gl...@uc...>wrote:

> Hello,
>
> Some of these functions may still have a few bugs.
>
> You can help debug them by first verifying that you can interpolate
> between two equal tensors, and then modifying the endpoints in
> different ways (same shape & different orientation, different shape &
> same orientation, different shape & different orientation, etc).
>
> If you can isolate what situations led to infinities then we will be
> in a better position to fix the code.
>
> Gordon
>
> On Nov 16, 2009, at 7:06 PM, Fangxiang Jiao wrote:
>
> > Hi Gordon & All,
> >
> > Thanks for all of the replies about my tensor interpolation question!
> >
> > Right now, I define a SCIRun module to call the tensor interpolation
> > functions.
> > The following is how I call the function:
> >
> >    int ptype=0;
> >
> >    if (check_option("method","linear"))
> >    { // default is linear
> >        ptype = tenInterpTypeLinear;
> >    }
> >    else if (check_option("method","geodesic-loxodrome-k"))
> >    {
> >        ptype =  tenInterpTypeQuatGeoLoxK;
> >    }
> >    else if (check_option("method","geodesic-loxodrome-r"))
> >    {
> >        ptype =  tenInterpTypeQuatGeoLoxR;
> >    }
> >    else if (check_option("method","riemannian"))
> >    {
> >        ptype = tenInterpTypeAffineInvariant;
> >    }
> >      // this last case is the Log-Euclidean
> >    else
> >    {
> >        ptype = tenInterpTypeLogLinear;
> >    }
> >
> >    double tenIntp[7];
> >    double tenA[7], tenB[7];
> >    tenA[0]=1;
> >    tenA[1]=1;
> >    tenA[2]=2;
> >    tenA[3]=3;
> >    tenA[4]=8;
> >    tenA[5]=6;
> >    tenA[6]=4;
> >
> >    tenB[0]=1;
> >    tenB[1]=3;
> >    tenB[2]=5;
> >    tenB[3]=4;
> >    tenB[4]=8;
> >    tenB[5]=3;
> >    tenB[6]=1;
> >    tenB[7]=1;
> >
> > tenInterpParm *tip = tenInterpParmNew();
> > tenInterpTwo_d(tenIntp, tenA, tenB, ptype, 0.5, tip);
> >
> > For the geodesic-loxodrome-k or geodesic-loxodrome-r, it works fine.
> > But
> > for the riemannian,
> > I got nans, and for the Log-Euclidean, I got infs. Do you know why it
> > didn't work?
> > I guess I may need to set the last parameter tip in some way, is
> > that right?
> >
> > Thank you so much!
> >
> > Fangxiang
> >
> >
> >
> >
> ------------------------------------------------------------------------------
> > Let Crystal Reports handle the reporting - Free Crystal Reports 2008
> > 30-Day
> > trial. Simplify your report design, integration and deployment - and
> > focus on
> > what you do best, core application coding. Discover what's new with
> > Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> > _______________________________________________
> > teem-users mailing list
> > tee...@li...
> > https://lists.sourceforge.net/lists/listinfo/teem-users
> >
>
>
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
> trial. Simplify your report design, integration and deployment - and focus
> on
> what you do best, core application coding. Discover what's new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> _______________________________________________
> teem-users mailing list
> tee...@li...
> https://lists.sourceforge.net/lists/listinfo/teem-users
>