R: bug in ARTKP? NaN errors from modelview matrix

How did you both compile?

I also got nan's when I compiled without "_USE_DOUBLE_"...

Cheers,

  Toni

> -----Messaggio originale-----
> Da: owner-artoolkit@h .................. [mailto:owner-
> artoolkit@h ..................] Per conto di Daniel Wagner
> Inviato: marted=EC 2 maggio 2006 14.27
> A: ARToolkit
> Cc: Tijs de Kler; Raymond de Vries
> Oggetto: Re: bug in ARTKP? NaN errors from modelview matrix
>=20
> Raymond de Vries wrote:
> > Attached are the used code and 2 images (640x480 raw) that are
> > generating pose estimations containing NaNs.
> >
> hi Raymond,
>=20
> I tested the code (basically the simple example that comes with
> ARToolKitPlus) with the images you attached to your emails. Here are =
the
> results i get:
>=20
> testImageNaN1.raw:
>   0.90  -0.07  -0.42  0.00
>   0.06  -0.96  0.29  0.00
>   -0.42  -0.28  -0.86  0.00
>   45.81  28.94  622.35  1.00
>=20
> testImageNaN2.raw:
>   0.94  -0.06  -0.35  0.00
>   0.07  -0.94  0.32  0.00
>   -0.35  -0.33  -0.88  0.00
>   50.75  27.56  645.26  1.00
>=20
> no NaN and the results look actually very reasonable (looking at the
> last rows...).
>=20
> In the sample code you sent me you load "LogitechPro4000.dat". Unless
> you really use that same camera file from the ARToolKitPlus =
distribution
> (which is btw completely wrong for a wide angle dragonfly lens) i see =
no
> reason why it would work on my machine and not on yours. I tested your
> files with ARToolKitPlus 2.1, but I doubt that i removed bugs in there
> since version 2.0.2 which would prevent those NaN.
>=20
> Any ideas?
>=20
> Daniel

From:	"Antonio Bleile" <antonio.bleile@s ........>	Received:	May 2, 2006
To	"'ARToolkit'" <artoolkit@h ..................>
Subject:	R: bug in ARTKP? NaN errors from modelview matrix
How did you both compile? I also got nan's when I compiled without "_USE_DOUBLE_"... Cheers, Toni > -----Messaggio originale----- > Da: owner-artoolkit@h .................. [mailto:owner- > artoolkit@h ..................] Per conto di Daniel Wagner > Inviato: marted=EC 2 maggio 2006 14.27 > A: ARToolkit > Cc: Tijs de Kler; Raymond de Vries > Oggetto: Re: bug in ARTKP? NaN errors from modelview matrix >=20 > Raymond de Vries wrote: > > Attached are the used code and 2 images (640x480 raw) that are > > generating pose estimations containing NaNs. > > > hi Raymond, >=20 > I tested the code (basically the simple example that comes with > ARToolKitPlus) with the images you attached to your emails. Here are = the > results i get: >=20 > testImageNaN1.raw: > 0.90 -0.07 -0.42 0.00 > 0.06 -0.96 0.29 0.00 > -0.42 -0.28 -0.86 0.00 > 45.81 28.94 622.35 1.00 >=20 > testImageNaN2.raw: > 0.94 -0.06 -0.35 0.00 > 0.07 -0.94 0.32 0.00 > -0.35 -0.33 -0.88 0.00 > 50.75 27.56 645.26 1.00 >=20 > no NaN and the results look actually very reasonable (looking at the > last rows...). >=20 > In the sample code you sent me you load "LogitechPro4000.dat". Unless > you really use that same camera file from the ARToolKitPlus = distribution > (which is btw completely wrong for a wide angle dragonfly lens) i see = no > reason why it would work on my machine and not on yours. I tested your > files with ARToolKitPlus 2.1, but I doubt that i removed bugs in there > since version 2.0.2 which would prevent those NaN. >=20 > Any ideas? >=20 > Daniel

From:	Daniel Wagner <daniel@i ...............>	Received:	May 2, 2006
To	ARToolkit <artoolkit@h ..................>
Subject:	Re: R: bug in ARTKP? NaN errors from modelview matrix
Antonio Bleile wrote: > How did you both compile? > > I also got nan's when I compiled without > "_USE_DOUBLE_"... > i simply used the settings from the simple example - nothing special. here's the output of the compile-time information: ARToolKitPlus v2.1: built May 2 2006 14:20:34 (MS C++ v14.0); floating-point; single precision; 6x6 marker; RGB pixelformat; custom memory manager; RPP support available. it does not work though with fixed point math... did you try using the RPP pose estimator? Daniel

From:	Raymond de Vries <raymond@s ......>	Received:	May 3, 2006
To	Antonio Bleile <antonio.bleile@s ........>
Subject:	Re: R: bug in ARTKP? NaN errors from modelview matrix
Hi Toni, I have been trying your suggestion to use doubles instead of floats (still in version 2.0.2) by defining _USE_DOUBLE_ (I needed to change TrackerImpl.h btw). Now the compile time info says: ARToolKitPlus compile-time information: ARToolKitPlus v2.0: built May 3 2006 11:30:22 (GCC 4.0.3); double precision; 6x6 marker; LUM pixelformat; RPP support available. But still the returned pose matrix contains NaNs... Ok, I will go to version 2.1.0 too and see what happens. Cheers Raymond Antonio Bleile wrote: >How did you both compile? > >I also got nan's when I compiled without "_USE_DOUBLE_"... > >Cheers, > > Toni > > > >>-----Messaggio originale----- >>Da: owner-artoolkit@h .................. [mailto:owner- >>artoolkit@h ..................] Per conto di Daniel Wagner >>Inviato: martedė 2 maggio 2006 14.27 >>A: ARToolkit >>Cc: Tijs de Kler; Raymond de Vries >>Oggetto: Re: bug in ARTKP? NaN errors from modelview matrix >> >>Raymond de Vries wrote: >> >> >>>Attached are the used code and 2 images (640x480 raw) that are >>>generating pose estimations containing NaNs. >>> >>> >>> >>hi Raymond, >> >>I tested the code (basically the simple example that comes with >>ARToolKitPlus) with the images you attached to your emails. Here are the >>results i get: >> >>testImageNaN1.raw: >> 0.90 -0.07 -0.42 0.00 >> 0.06 -0.96 0.29 0.00 >> -0.42 -0.28 -0.86 0.00 >> 45.81 28.94 622.35 1.00 >> >>testImageNaN2.raw: >> 0.94 -0.06 -0.35 0.00 >> 0.07 -0.94 0.32 0.00 >> -0.35 -0.33 -0.88 0.00 >> 50.75 27.56 645.26 1.00 >> >>no NaN and the results look actually very reasonable (looking at the >>last rows...). >> >>In the sample code you sent me you load "LogitechPro4000.dat". Unless >>you really use that same camera file from the ARToolKitPlus distribution >>(which is btw completely wrong for a wide angle dragonfly lens) i see no >>reason why it would work on my machine and not on yours. I tested your >>files with ARToolKitPlus 2.1, but I doubt that i removed bugs in there >>since version 2.0.2 which would prevent those NaN. >> >>Any ideas? >> >>Daniel >> >> > > > > -- Raymond de Vries, VR Specialist Insight Services SARA Computing and Networking Services, PO Box 94613, 1090 GP AMSTERDAM email: raymond at sara.nl Phone +31-20-5923000 Fax +31-20-6683167