Artk+

i have this old artk+ discussion in my browser. dated back from Jun03:

********************==
sonderformat: [03.06.07 03:31 ](03.06.07 03:31 )

@foxonfire

you wrote, that you copied the ArToolKitPlus.dll of the artk+ distribution into the vvvv bin. this is totally the wrong way.

you don’t really need the “ARToolkitPlus library”.
only to get markers and a camera calib file.

just download the ARTK+Tracker.zip.
unzip the ARTK+Tracker.dll and the helpfile into the vvvv “freeframe” folder and start the helppatch.

if it is not working, what exactly is not working?
do you get a camera image?

foxonfire: [04.06.07 11:14 ](04.06.07 11:14 )

sonderformat:

i have finally get it working after the weekend !!!

thanks for your detailed steps which made me looking into everything again, it was my fault copying a bad camera calibration file to the Freeframe folder… so many .dat product when i was playing around with ARTK+ previously :P

now i can see a good list of tag numbers when the markers show up. Nice!!!

  • absolutely no need to copy the ARTookKitPlus.dll to the path. (it was one of the desperate moves i tried when there was no clue at all…)

  • camera calib file is essential: use the one supplied with the ARTK+ package is a good start (LogitechPro4000.dat). it should work with most webcams (i tried)

  • if you suspect the webcam is not working, you can replace the ‘videoin’ with a ‘filetexture’->‘asVideo’ and load a test image from the ARTK+ (e.g. image_320_240_8_marker_id_bch_nr0100.jpg)

hope this is useful in case someone run into problem.

dujoducom: [05.06.07 04:59 ](05.06.07 04:59 )

I can’t get this to work, where do you get the .pat files for use with the “marker” node? I’ve found the patterns to print, they are saved as .png in the id-markers/simple/ folder, but I can’t find the .pat files (I just searched for *.pat). Am I missing something?

tonfilm: [05.06.07 05:07 ](05.06.07 05:07 )

don’t you mean .dat?

dujoducom: [05.06.07 05:13 ](05.06.07 05:13 )

no, the camera file I can find without a problem. It’s the .pat files that are apparently supposed to be loaded into the pin labeled “Marker Files” that I can’t find.

joreg: [05.06.07 06:53 ](05.06.07 06:53 )

you don’t need a .pat file. they are only needed for custom patterns but that functionality is not implemented yet. all the markers you found for printing are detected automatically.

foxonfire: [05.06.07 10:12 ](05.06.07 10:12 )

a quick question on ‘marker width’: i used to set the ‘border width’ in ARTK+ (for bch-thin it is 0.125). is that correct if i put 0.75 for ‘marker width’ in this case?

joreg: [05.06.07 16:01 ](05.06.07 16:01 )

thanks, good question. forgot to mention that:
the value is to be given in millimeters/100. the default is 80mm/100 = 0.8. it means the actual physical width of your markers.

elektromeier: [06.06.07 16:13 ](06.06.07 16:13 )

hello

i have some troubles with the stability of the tracker…
i observed two problems:

* the markers get lost completly often for one frame (echanged with a -1 id) 


* the id stays correctly but the transform values change very rapidly to very high values, so high i cant filter them with damper anymore 

i tested with a bw cctv camera with variable zoom and iris. how can i improve the stabilty? could that instability be caused of the wrong cameracalibration file (LogitechPro4000.dat)? what cameraparams are defined in that file?

cheers
ele

kalle: [06.06.07 21:30 ](06.06.07 21:30 )

hint:

if you are wondering about strange transformations of the teapot:
disconnect translate and rotate from the renderer (or reset them…)

sonderformat: [06.06.07 21:45 ](06.06.07 21:45 )

helo ele
yes.
the instability you describe is a big problem which
i have not solved till now.

i’m nearly shure that it is not the camera calib file.

Daniel Wagner from artk+ told me that the tracking must be filtered
by “Kalman” or “Desp” Filter to get more stable.
Kalman should be better but comparing to the effort of implementing DESP, the DESP should be the choice

don’t know if this could be done in VVVV.
should better be implemented in the Freeframe-Code i think but haven’t understood howto till now.

see here.

kalle: [06.06.07 23:38 ](06.06.07 23:38 )

btw.:

projection output of the tracker seems to be constant.
for what reason is the S+H (Animation) in the helpfile?

sonderformat: [07.06.07 00:31 ](07.06.07 00:31 )

@kalle
i think you’re right with the S+H-Node.

kalle, do you mean the strange transformations you get by using a damper? how and when do you reset the transforms to be aware of this?

oschatz: [07.06.07 00:37 ](07.06.07 00:37 )

@sonderformat: should better be implemented in the Freeframe-Code i think but haven’t understood howto till now.

no - better have it in one vvvv module, so it can be used with all different kind of trackers…

kalle: [07.06.07 04:55 ](07.06.07 04:55 )

@sonderformat:
with “strange transformations” i mean that the helpfiles “teapot” has an offset translation and rotation relative to the “marker”-quad.
this is due to the settings of translate and rotate node which are connected to the renderers projection transform inlet.
seems that the helpfile was packed in hurry…

today i read on skype that a least two other users tried to retransform that stuff on the teapot and spended some time on this…

another thing 'bout the marker size:

my markers are about 80mm²
and i am SURE that i had it running pretty well this afternoon (at work) with the recommended setting 80/100 = 0.8.

and now (at home) i have to set it to 0.0110 !!!
same camera.
and same result with girlfriend’s daughter’s webcam.

sonderformat: [07.06.07 21:11 ](07.06.07 21:11 )

marker size
sorry kalle. don’t know why you have had to down to 0.0110 marker width.
80mm² are quasi 9*9mm says tuniel.
this physical marker width/100 is working pretty good for us.
also with totally different cams.

transformation
i thought you always have to change the transformations
depending by the size of the model you display on the marker.
and it has also to be pitched by 0.25

when i think about it at the moment why not doing this dynamically?

damping
i tried to damp the tracking. but i always get this strange behaviour
(see gaultier screenshot).
i’m wearing this shirt to put artoolkit tracking to the acid test. harhar!

see this. i tried to damp with dynamic filtertime.
i think this is not the ultimate truth in this case.

attachment artk+damper.zip (3.91 Kb)

Desaxismundi: [08.06.07 18:58 ](08.06.07 18:58 )

Here, u can find the Help patch revisited (tx 2 Elektromeier)…it works for me…Hope it helps!!

attachment ARTK+Tracker.zip (513.98 Kb)

Thanks for reposting this information. Do you (or does anyone else) happen to have the revisited help patch Desaxismundi? I’m trying again to get this to work and I can get the IDs of the markers recognized, but I still can not get the teapot (or anything else) to align with the markers.

Here it is!!!..but undocumented.

ARTK+Tracker.zip (526.6 kB)

Ah, thanks a lot, that one definitely works!

i have problems to get this working. vvvv recognizes the markers, but gives always a negative value (-1). anybody have any idea what could be causing this. my webcam is logitech 4000 and i’m using the corresponding .dat file.

The patch above should do it as well…tried it already?

yep. tried it alright. the result is the same. and by the way thanks for the patch. this and ARTag are the only ones I’ve got to work at least somehow. i tried diffrent patterns as well and playing around with the pattern width, but no result. always -1.

manually changing the ID pin to something else than -1 works, but i know it’s a bad fix.

http://studierstube.icg.tu-graz.ac.at/handheld_ar/images/artoolkitplus_BCH_thin_board.png

It’s buggy to say the least.
Is there any one else getting a 10 marker limit?
Starting to track in the upper left corner and then stops when it hits around 10 markers. If I shadow the marked markers it will select the next couple of markers it finds.

Also it scales things to insane values when it jumps around.
I get -1 a lot also. But I usually get a stable ID if I tilt the marker. Not sure why though.

I’m finding it impossible to get custom camera calibration data working with this node.
The ARTK+ website says that it accepts ASCII files using variables calculated in a matlab program. this node doesnt seem to load these files though

the ones attached by desaxismundi are binary files and they work

the ones created by the standard ARToolKit calibration program are binary (and look similar to the ones which come with desaxismundi’s helpfile) but this node wont load them

i’m pulling my hair out :’(
how do i calibrate this node for my 3.5mm uEye camera?

As well as the calibration camera
Has anyone had any success at all tracking any markers? I always get ID of -1 and i’ve tried markers from what seems like almost everywhere (hair comes out some more)

yes, it works quite well for me. the limit of markers is set somewhere in the freeframe plugin, i had to recompile it with code::blocks, now it tracks up to 16 markers.
but i was not using any custom calibration file…

ARTK+Tracker.dll (257.5 kB)

Good news on the markers
any advice on how to calibrate a camera? I’ve used the GLM tools (standalone and MATLAB) and the ARToolkit one but cant find anything that’ll make a file which ARTK node will understand

I was playing around with this wonderfull thing now for some time.
Worked well for me, the flickering didn’t realy bother me. (by the way: thanks for this thread)
But now as i started to get an idea of what i could do… this thing stopped working.

I’m still getting the IDs of the tags, but the ModelView Values are more than flickering… they seem to be totally random. (had similar results befor, but then just for a few seconds)

Can this be connected to some other software on the pc. I remember installing the new iTunes 8, but I’m not sure if this was exactly the time it stopped working. (tried to uninstall it though) I also and install the newest DX9.24 and removed vvvv and started the artk-tracker thing all over… but no change anymore to get any result.

Has someone an idea what i can do, or what might be wrong?
I’ve no idea what else I could do.

Thanks, M.

hmmm, strange results there.
It seems to be a pretty strange node all in all.

Any news on whether it’s possible to make your own custom calibration files for this node? and if so, with what tool? (i’ve got Matlab if needs be!)

If not, can anyone suggest a camera that will work well for tracking 15cm ID’s from about 4m away?
It’s going to be in a club environment with lots of lights going off, so I’ll need to work in IR with an illuminator and make sure to not hire any lights that kick out IR (i.e. LED’s only)

@sugoku: so you have tried everything mentioned on the very bottom of this page?!

can you post one of your generated camerafiles?