cap10subtext

https://github.com/cap10subtext/jit.artkmulti

I'm just totally swallowing my pride on this one and posting this obscene mess of code. Forgive the cliche but I'm NOT a classically trained programmer, it's been my one of my pet projects for learning C which of course means lots of trial and error and beginners mistakes.

In particular I'm having issue with deleting and creating a new object and changing the camera dims during run time.

One thing that's missing from this code post (though it should be obvious what's required), you need a folder called artk.support in the Max5/ folder (can't be in the cycling 74 folder at the moment, fopen rejects any spaces in the path to camera.dat or multi/objectdata2

objectdata2 is being used with 4 obejcts, currently hard coded for 4 outlets.

The help patch is particularly scary but it's just an early version I picked out at this late hour so there was something to test with.

I know it's rough but I've heard a lot of interest in this recently. Hopefully someone will find it useful.

New Project on github.
https://github.com/cap10subtext/jit.artkmulti

I'm just totally swallowing my pride on this one and posting this obscene mess of code.  Forgive the cliche but I'm NOT a classically trained programmer, it's been my one of my pet projects for learning C which of course means lots of trial and error and beginners mistakes.

I know it's rough but I've heard a lot of interest in this recently.  Hopefully someone will find it useful.

Currently the compile is Intel Mac only.


artoolkit-for-maxmsp

jit.artkmulti: unable to load object bundle executable

2011-03-06 10:46:56.953 MaxMSP[4444:20b] Error loading /Applications/Max5/artk.support/jit.artkmulti.mxo/Contents/MacOS/jit.artkmulti: dlopen(/Applications/Max5/artk.support/jit.artkmulti.mxo/Contents/MacOS/jit.artkmulti, 262): no suitable image found.

    /Applications/Max5/artk.support/jit.artkmulti.mxo/Contents/MacOS/jit.artkmulti: mach-o, but wrong architecture

Anyway, this going to be very nice to artoolkit for MAx ! Just have to make it work :)

Jitter 1.7.0 installed
jit.artkmulti: unable to load object bundle executable
 2011-03-06 10:46:56.953 MaxMSP[4444:20b] Error loading /Applications/Max5/artk.support/jit.artkmulti.mxo/Contents/MacOS/jit.artkmulti:  dlopen(/Applications/Max5/artk.support/jit.artkmulti.mxo/Contents/MacOS/jit.artkmulti, 262): no suitable image found.
 Did find:
 	/Applications/Max5/artk.support/jit.artkmulti.mxo/Contents/MacOS/jit.artkmulti: mach-o, but wrong architecture

Anyway, this going to be very nice to artoolkit for MAx ! Just have to make it work :)
Cheers.
Ad


Will check compiler settings. What type of machine are you on?

Will check compiler settings.  What type of machine are you on?


but I have a strange behavior in my patch. the cube (or whatever it's rendered) is not erased when loosing the track of the pattern. It's supposed to be like that? A work around would be to send a not draw (or something similar) message to jit.gl.sketch or jit.gl.model when the pattern is lost, but for doing so some index number (0 or 1) should be sent together with the list with coordinates. Or we could do something else to get the object erased?

great work!
but I have a strange behavior in my patch. the cube (or whatever it's rendered) is not erased when loosing the track of the pattern. It's supposed to be like that? A work around would be to send a not draw (or something similar) message to jit.gl.sketch or jit.gl.model when the pattern is lost, but for doing so some index number (0 or 1) should be sent together with the list with coordinates. Or we could do something else to get the object erased?


Ad.: I've tweaked something in the compiler that may have been causing the problem. Please feel free to try this one.

sandroid: the first element of the list is now 1 or 0 (visible or not) so you can connect your gl object to an enable $1 message to add or remove your model from the scene.

https://github.com/downloads/cap10subtext/jit.artkmulti/ARTookitMax.zip

Ad.: I've tweaked something in the compiler that may have been causing the problem.  Please feel free to try this one.  

New Downloads are here:
https://github.com/downloads/cap10subtext/jit.artkmulti/ARTookitMax.zip


Damn, I just typed out this big, detailed post and my browser ate it....

Take 2... (greatly summarized due to frustration of having to retype it all):

Here are the main issues I'm contending with:

1. Including external files: fopen won't accept (even properly escaped) spaces. Suggestions? Where to keep file, workarounds to fopen?

2. Can you realloc after jit_object_alloc? I've checked that the sizeof everything I want to realloc into x:

x = (t_jit_artoolkit *)realloc(x, sizeof(x) + (sizeof(ObjectData_T) * *objnum));

seems to be within the 16384 bytes limit so I'm a little confused (not very savvy with memory management in C).

3. Reinit based on change of incoming dims: obviously Jitter handles this a little differently than ARToolkit. My attempt to do this:

        if ((in_minfo.dim[0]!= xsize)||(in_minfo.dim[1]!= ysize)) {

            arVideoCapStop();

            xsize = in_minfo.dim[0];

            ysize = in_minfo.dim[1];

            post("input dimensions changed");

            jit_artkmulti_param_load();

Sends ARToolkit into a death spiral. So clearly I'm not re-initializing ARToolkit properly. Perhaps someone familiar with ARToolkit will stumble across this post and point out my silly mistake.

Damn, I just typed out this big, detailed post and my browser ate it....
Take 2... (greatly summarized due to frustration of having to retype it all):

Here are the main issues I'm contending with:
1. Including external files: fopen won't accept (even properly escaped) spaces.  Suggestions?  Where to keep file, workarounds to fopen?
2.  Can you realloc after jit_object_alloc?  I've checked that the sizeof everything I want to realloc into x:
x = (t_jit_artoolkit *)realloc(x, sizeof(x) + (sizeof(ObjectData_T) * *objnum));
seems to be within the 16384 bytes limit so I'm a little confused (not very savvy with memory management in C).
3.  Reinit based on change of incoming dims: obviously Jitter handles this a little differently than ARToolkit.  My attempt to do this:
		if ((in_minfo.dim[0]!= xsize)||(in_minfo.dim[1]!= ysize)) {
			arVideoCapStop();
			arVideoClose();
			argCleanup();
			xsize = in_minfo.dim[0];
			ysize = in_minfo.dim[1];
			post("input dimensions changed");
			jit_artkmulti_param_load();
		}
Sends ARToolkit into a death spiral.  So clearly I'm not re-initializing ARToolkit properly.  Perhaps someone familiar with ARToolkit will stumble across this post and point out my silly mistake.


despite the problems you just related, I's really cool to have artoolkit ported to jitter.

I will use a lot with my students this year.

thanks for the update!
despite the problems you just related, I's really cool to have artoolkit ported to jitter.
I will use a lot with my students this year.

New Project page up (easier download link for those scared of github) ;)

New Project page up (easier download link for those scared of github)   ;)

Apparently you made the exrernal with xcode 4, 10.6 leopard…

that might be the reason it doesn't work on 10.5.

Hello
Apparently you made the exrernal with xcode 4, 10.6 leopard…
that might be the reason it doesn't work on 10.5.

Ad.: Xcode 3.2.5 actually, with the base system being set to 10.5. I tested an earlier version on my other computer which is 10.5. Will do again and see if I can find the problem. Just been a bit busy lately.

Ad.: Xcode 3.2.5 actually, with the base system being set to 10.5.  I tested an earlier version on my other computer which is 10.5.  Will do again and see if I can find the problem.  Just been a bit busy lately.


Okay, I've confirmed it's not a leopard issue, this is running on my leopard machine. Try downloading the new version and getting rid of any existing ones.

Okay, I've confirmed it's not a leopard issue, this is running on my leopard machine.  Try downloading the new version and getting rid of any existing ones.

oups— you're right… I reinstal max, but nothing change…

Can anyone reproduce the bug ?
Just dont get it…
mac book pro 10.5.8.
Max 5.1.7
Java 1.5
Ad-


Can you replicate what I've just done and post a screen shot with object, Max printout, and machine info? Maybe it'll give me a clue...

Can you replicate what I've just done and post a screen shot with object, Max printout, and machine info?  Maybe it'll give me a clue...


Ad. I've posted a new download but I'm not sure if it'll help. Be absolutely sure you've trashed all the old versions and then put artk.support at /Max5.

It's compiled as 32-bit Universal, Base system 10.5, valid architectures x86_64 i386. That should do the trick, and I've ensured that the project settings are the same for the executable as they are for the project info.

If that's not it, we may need help from the c74 crew on this...

Ad. I've posted a new download but I'm not sure if it'll help.  Be absolutely sure you've trashed all the old versions and then put artk.support at /Max5.

(It's at the same link)
http://www.aranarproductions.com/artk/

It's compiled as 32-bit Universal, Base system 10.5, valid architectures x86_64 i386.  That should do the trick, and I've ensured that the project settings are the same for the executable as they are for the project info.

If that's not it, we may need help from the c74 crew on this...


I format my entire disc, and everything is just fine now !

I format my entire disc, and everything is just fine now !
Thank you.
Ad.


Glad it's working! Sorry you had to go through all that trouble, but all the same, I'm glad it's not the external causing you grief. ;)

Glad it's working!  Sorry you had to go through all that trouble, but all the same, I'm glad it's not the external causing you grief.  ;)


My Idea was to connect artoolkit with the jit.ogre external, but it seems there is a bug when i try to connect jit.ogre models and draw_pixels.

My jit.window is showing the both but flash...

Hello cap10subtext,
thanks a lot for this external,
My Idea was to connect artoolkit with the jit.ogre external, but it seems there is a bug when i try to connect jit.ogre models and draw_pixels.
My jit.window is showing the both but flash...
do you have any suggestions?


Hi Pulsart, check out the following post, I've had some issues with the jit.ogre external too (the potential is irresistible, no?) which are outlined there (including the draw_pixels issue and an animation "bug?" I'm sure you'll come across). Rob posted a solution with a videoplane that I found worked in some situations but not in others. So I'm not sure what the ideal solution is I had a few issues with it regarding mixing it with other GL shapes (his solution is posted in that thread). Maybe it'll work for your purposes?

https://cycling74.com/forums/jit-ogre-for-max-5

The other thing I was having trouble with was changing the z offset and rotation transform on jit.ogre objects. But perhaps mention to Rob (or add to the jit.ogre thread) that you too would be interested in seeing if an equivalent to the gltranslate or glrotate feature could be added to the jit.ogre object?

Hi Pulsart, check out the following post, I've had some issues with the jit.ogre external too (the potential is irresistible, no?) which are outlined there (including the draw_pixels issue and an animation "bug?" I'm sure you'll come across).  Rob posted a solution with a videoplane that I found worked in some situations but not in others.  So I'm not sure what the ideal solution is I had a few issues with it regarding mixing it with other GL shapes (his solution is posted in that thread).  Maybe it'll work for your purposes?

The other thing I was having trouble with was changing the z offset and rotation transform on jit.ogre objects.  But perhaps mention to Rob (or add to the jit.ogre thread) that you too would be interested in seeing if an equivalent to the gltranslate or glrotate feature could be added to the jit.ogre object?


I am very interested too in the possibility of integrating Artoolkit and Jit.ogre.

https://github.com/downloads/cap10subtext/jit.artkmulti/jit.artkmulti-PC.zip

This is the first compile I've done for ARToolkit for Max/MSP/Jitter for the PC! So far tested on Windows XP. I need PC Beta testers since I have limited access to a PC. Anyone who can help me with the code is welcome to as well, there are few differences between this and the Apple compile notably the path names and a few different headers. Could also use some pointers about the resolution. 640x480 is fine on a Mac for tracking markers, it seems really pixellated on the Windows machines I've tried.

There were a lot of issues with this code, but instructions are all there in the README to get it working. There was a huge issue with the dlls but the ones you need are included, simple install them in the local support folder. Again I have to remind all that the code is provided AS-IS. I bear no responsibility for anything that might happen to your system as a result of using my terrible code. :)

Note the other thing different about this version is that you need to install the multi folder in your local project folder before you start your jit.artkmulti object. Otherwise it won't find the markers and you'll get a crash when you start the video.

This is the first compile I've done for ARToolkit for Max/MSP/Jitter for the PC!  So far tested on Windows XP.  I need PC Beta testers since I have limited access to a PC.  Anyone who can help me with the code is welcome to as well, there are few differences between this and the Apple compile notably the path names and a few different headers.  Could also use some pointers about the resolution.  640x480 is fine on a Mac for tracking markers, it seems really pixellated on the Windows machines I've tried.

There were a lot of issues with this code, but instructions are all there in the README to get it working.  There was a huge issue with the dlls but the ones you need are included, simple install them in the local support folder.  Again I have to remind all that the code is provided AS-IS.  I bear no responsibility for anything that might happen to your system as a result of using my terrible code.  :)

Note the other thing different about this version is that you need to install the multi folder in your local project folder before you start your jit.artkmulti object.  Otherwise it won't find the markers and you'll get a crash when you start the video.


Hi cap10subtext - a quick bug report on windows 7 (not sure if its just me or everyone though), the object looks for camera_para and object_data files in C:Program FilesCycling '74Max 5.0artk.support , although for most people it will be in C:Program Files (x86)Cycling '74Max 5.0artk.support

I checked I have the paths set up correctly (with the (x86)) in preferences. I overcame the problem by copying artk.support to a new path where it was looking for the files.

Hi cap10subtext - a quick bug report on windows 7 (not sure if its just me or everyone though), the object looks for camera_para and object_data files in C:Program FilesCycling '74Max 5.0artk.support , although for most people it will be in C:Program Files (x86)Cycling '74Max 5.0artk.support 

I checked I have the paths set up correctly (with the (x86)) in preferences. I overcame the problem by copying artk.support to a new path where it was looking for the files.


Thanks for the report. If anyone else has Windows 7 (64 bit) I'd appreciate knowing if that path is uniform between machines.

Thanks for the report.  If anyone else has Windows 7 (64 bit) I'd appreciate knowing if that path is uniform between machines.


New download posted on Github to correct path name to C:Program Files (x86)Cycling '74Max 5.0artk.support for anyone else with this issue. Would love feedback as to whether this solves the problem.

Note: it's not a 64 bit compile, it's just the different path name.

New download posted on Github to correct path name to C:Program Files (x86)Cycling '74Max 5.0artk.support for anyone else with this issue.  Would love feedback as to whether this solves the problem.

Note: it's not a 64 bit compile, it's just the different path name.


Is it possible to organise artk to follow the standard max paths?

I generally keep all my foreign objects in a particular folder which is added to MaxPath.

Artk did not work from here and seemed to want to be in Max5 root. Would prefer not to symlink . .

Apart from that it seems to work. Mac i7 QC, 10.6.

Hi cap10,
Is it possible to organise artk to follow the standard max paths?
I generally keep all my foreign objects in a particular folder which is added to MaxPath.
Artk did not work from here and seemed to want to be in Max5 root. Would prefer not to symlink . .
Apart from that it seems to work. Mac i7 QC, 10.6.

Macciza: May not be possible, there's an issue with the function that parses the path name that won't allow for spaces in the paths on a Mac and I think that it's in one of the libs. Have had a hard time tracking it down and don't know if I can recompile it.

Would also need an example of how to use the Max paths in an external.

Macciza: May not be possible, there's an issue with the function that parses the path name that won't allow for spaces in the paths on a Mac and I think that it's in one of the libs.  Have had a hard time tracking it down and don't know if I can recompile it.

Okay, after tinkering a bit I think I've solved the issue with spaces in the path name. Turns out it was a parsing issue in one of the classes.

If i can figure out how using a max path works for other externals I may be able to implement it after all.

Okay, after tinkering a bit I think I've solved the issue with spaces in the path name.  Turns out it was a parsing issue in one of the classes.

If i can figure out how using a max path works for other externals I may be able to implement it after all.


New object posted for OS X, should work with the Max File paths now. Just place the folder anywhere in the Max search paths and add this:

max fileformat .dat DATA 0 "data file" file;

/Applications/Max5/Cycling '74/init/max-fileformats.txt

Request to Cycling '74 to add the Data file to max-fileformats.txt permanently.

To build a standalone, just add the camera_para.dat and the multi folder to .app/Contents/support/

https://github.com/downloads/cap10subtext/jit.artkmulti/ARToolkitMax.zip

This release works with a new format for object_data2 in /multi, so you should use the new one with the new object.

New object posted for OS X, should work with the Max File paths now.  Just place the folder anywhere in the Max search paths and add this:

This release works with a new format for object_data2 in /multi, so you should use the new one with the new object.


Hi cap10subtext - I've tested your win7 x64 fix and it works fine for me (I have deleted the old non-(x86) folder).

I also have a question - in the windows readme, does the line "The calibration of the camera is default .dat camera file only" mean its not possible to replace the default camera_para.dat file with a user-made one?

It's possible, but may not be desirable at this time. I tried a different calibration that was MacBook iSight specific and instead of improving tracking it mucked it up. Not sure what the fix for that would be at this time. If someone who knew ARToolkit front to back could point out the issue in the code I'd fix it in a heartbeat...

It's possible, but may not be desirable at this time.  I tried a different calibration that was MacBook iSight specific and instead of improving tracking it mucked it up.  Not sure what the fix for that would be at this time.  If someone who knew ARToolkit front to back could point out the issue in the code I'd fix it in a heartbeat...


Interesting. Maybe try a test dat after doing the one-step calibration yourself: I'm using a video camera (HV20) + capture stick and found it reduced flickering quite a lot (although I also changed some lighting around that time too, so I couldn't say for sure it was just calibration).

Also another note - on your website it says to edit object_data2, but so far mine has only been reading object_data (I've actually renamed 2 to avoid confusion). I am only using single marker tracking (of 4 markers) though, so if thats related to multi-marker tracking, ignore this.

Thanks for producing this external - its awesome and just what I need for a project I'm doing in a nail salon soon. The only challenge at the moment is how temperamental artoolkit is - every morning I load up the working patch from last night and have to find out why its now crashing.

But this seems to be the way with the standalone artoolkit binaries too (crashing if the camera is the wrong kind, or the marker file isn't found). I guess its the price of working with such specialist tools.

Thanks for producing this external - its awesome and just what I need for a project I'm doing in a nail salon soon. The only challenge at the moment is how temperamental artoolkit is - every morning I load up the working patch from last night and have to find out why its now crashing. 

But this seems to be the way with the standalone artoolkit binaries too (crashing if the camera is the wrong kind, or the marker file isn't found). I guess its the price of working with such specialist tools.


I just checked and the alignment is in fact better with the default file. (Assuming that the dat file doesn't influence the tracking, and only affects the alignment of objects rendered.)

This could be because I failed to play with the lens_angle and position offsets to counter the new alignment, but to my eye the default is good enough for me.

Thanks for the kind words. It's been a dream of mine for the longest time to get something working natively in Max but of course it's come together finally now that I barely have time to work on it. So I apologize for the crashing, I just haven't had the time to to do all the error checking that the open source ARToolkit seems to lack to make it play nice with Max.

And I apologize in advance that the windows versions won't get quite the same level of attention because frankly I hate using windows. Would be more than happy to post the Visual Studio 2008 or 2010 project files if someone wants to help maintain it.

And for windows it is object_data not object_data2. As I said, not a lot of time to throw this together so my naming conventions are out of whack! :)

Thanks for the kind words.  It's been a dream of mine for the longest time to get something working natively in Max but of course it's come together finally now that I barely have time to work on it.  So I apologize for the crashing, I just haven't had the time to to do all the error checking that the open source ARToolkit seems to lack to make it play nice with Max.

And for windows it is object_data not object_data2.  As I said, not a lot of time to throw this together so my naming conventions are out of whack!  :)


Oh and regarding the calibration, it's because there's a bit of a hack involved to get a coordinate system that works with out too much fiddling in Max. In ARToolkit, I believe there is an extra step they do to apply the resulting projection Matrix from the camera file to the transformation matrix but it's done as a rendering step in OpenGL (I could be mistaken about certain details) which make calls to the glu viewport which we aren't using in Max so I just bypassed it and corrected the tracking to work with the default file. Ulitmately would love to have it work properly but I'd need to pick the brain of an ARToolkit expert and a better C programmer than I.

Oh and regarding the calibration, it's because there's a bit of a hack involved to get a coordinate system that works with out too much fiddling in Max.  In ARToolkit, I believe there is an extra step they do to apply the resulting projection Matrix from the camera file to the transformation matrix but it's done as a rendering step in OpenGL (I could be mistaken about certain details) which make calls to the glu viewport which we aren't using in Max so I just bypassed it and corrected the tracking to work with the default file.  Ulitmately would love to have it work properly but I'd need to pick the brain of an ARToolkit expert and a better C programmer than I.


New object posted for Windows, this object loads properly on my machine (XP SP 3) but I'm having all sorts of unrelated webcam issues and can't test it properly. Please let me know if it's not working for you.

This release should work with the Max File paths now, therefore the x86 path fix should be no longer necessary, please feel free to report any bugs. Just place the support folder anywhere in the Max search paths and REMEMBER to add this:

Program FilesCycling '74Max 5Cycling '74initmax-fileformats.txt (or the Runtime equivalent).

I'd like to request to Cycling '74 to add the Data file to max-fileformats.txt permanently, if anyone has any complaints or a conflicting standard please let me know.

Have never built a PC standalone but should work when dlls, camera params and multi folder copied to support folder.

https://github.com/downloads/cap10subtext/jit.artkmulti/ArtoolkitMaxPC.rar

This release works with a new format for object_data2 (note it no longer supports object_data from the original download) in multi. Please erase old support folders in your path to avoid a conflict.

New object posted for Windows, this object loads properly on my machine (XP SP 3) but I'm having all sorts of unrelated webcam issues and can't test it properly.  Please let me know if it's not working for you.

This release should work with the Max File paths now, therefore the x86 path fix should be no longer necessary, please feel free to report any bugs.  Just place the support folder anywhere in the Max search paths and REMEMBER to add this:

This release works with a new format for object_data2 (note it no longer supports object_data from the original download) in multi.  Please erase old support folders in your path to avoid a conflict.

Copy of GPL license added to PC release.


Hi cap10subtext - thanks for the update. I've had the previous version running without issues since a while after my last post, once I narrowed down some 'rules' to avoid crashes. Basically the bullet-proof way to launch it for me is by double-clicking the patch I want to edit while Max is closed. The fact opening it from the file menu in Max doesn't work makes me suspect it might be related to filepaths, so maybe this new version will fix that.

I've created a rough proof-of-concept of an automatic-thresholder for single black and white markers. It should reduce drop-outs due to changes in light like glare or shadows. I'm about to move to colour markers, so I won't have time to work further on it. I think a lot of the heavy lifting involved could be moved to the GPU, but at the moment it isn't too intensive.

It's very rough so I think the patching that determines how often it runs doesn't actually work properly (did it late at night) and the patching in general is just how it came out while working through the problem (this is the first patch I've posted).

Additional improvements could include running the 3m on each plane for colour markers (and only using full/zero RGB values on the markers like 255 0 0 / 255 255 0 /0 0 0), or running for a few frames after losing the marker and widening the sample radius (in case it can pick it up again in the same area with more contrast data). Note the patch is looking for marker 2.

I've created a rough proof-of-concept of an automatic-thresholder for single black and white markers. It should reduce drop-outs due to changes in light like glare or shadows. I'm about to move to colour markers, so I won't have time to work further on it. I think a lot of the heavy lifting involved could be moved to the GPU, but at the moment it isn't too intensive. 

Additional improvements could include running the 3m on each plane for colour markers (and only using full/zero RGB values on the markers like 255 0 0 / 255 255 0 /0 0 0), or running for a few frames after losing the marker and widening the sample radius (in case it can pick it up again in the same area with more contrast data). Note the patch is looking for marker 2.


"Basically the bullet-proof way to launch it for me is by double-clicking the patch I want to edit while Max is closed. The fact opening it from the file menu in Max doesn't work makes me suspect it might be related to filepaths, so maybe this new version will fix that."

Hmmm... interesting about the file menu issue, hadn't noticed that. Hopefully this does get resolved with the latest update.

Glad it's working out for you. The way I see it, there are likely 3 more very important issues, however, it's possible that they will never fully be resolved.

1) I need to figure out how to properly handle the memory of this object. I have an example which I'm waiting on which should give me an idea of how to accomplish this by properly allocating memory to the loading of the marker "objects".

2) There doesn't seem to be a way to safely handle the ARToolkit libraries, the normal cleanup routines don't have any effect.

3) there must be a way to correct for the proper camera calibration files but this totally escapes me at the moment. You can see when the marker gets further from the camera the skew becomes more obvious.

The final issue which is less pressing is that the "detection" output is still not precise... It seems to have trouble with multiple markers entering and leaving the scene simultaneously.

Having said all of that, I like where you are going with this patch (you are right in that it seemed to attempt the fix once, then it would stop working). Happy augmenting!

Hmmm... interesting about the file menu issue, hadn't noticed that.  Hopefully this does get resolved with the latest update.

Glad it's working out for you.  The way I see it, there are likely 3 more very important issues, however, it's possible that they will never fully be resolved.
1) I need to figure out how to properly handle the memory of this object.  I have an example which I'm waiting on which should give me an idea of how to accomplish this by properly allocating memory to the loading of the marker "objects".
2) There doesn't seem to be a way to safely handle the ARToolkit libraries, the normal cleanup routines don't have any effect.
3) there must be a way to correct for the proper camera calibration files but this totally escapes me at the moment.  You can see when the marker gets further from the camera the skew becomes more obvious.

Having said all of that, I like where you are going with this patch (you are right in that it seemed to attempt the fix once, then it would stop working).  Happy augmenting!


I'm trying to move from single markers to multiple markers now (tracking multiple independent object, not single-position multi-pattern markers), but I'm not 100% sure if the object is supposed to support this.

At the moment I have 4 patterns loaded, but when multiple patterns are on screen, the object will only output position data for one. If I cover that marker up, it will switch to outputting another. Is the multi-marker tracking only for the kind of multi-pattern, single point tracking shown here: 

http://www.hitl.washington.edu/artoolkit/documentation/tutorialmulti.htm

 or is something wrong in my objectdata formatting etc.?

Additionally, I realised I was needlessly recreating cv.jit.threshold above, so I've switched to that for much more stable marker detection in changing lighting conditions. I'd like to find a shader that does something similar, as all the cv objects can slow things down a lot, but I've yet to come across one other than a pixelbender script.

At the moment I have 4 patterns loaded, but when multiple patterns are on screen, the object will only output position data for one. If I cover that marker up, it will switch to outputting another. Is the multi-marker tracking only for the kind of multi-pattern, single point tracking shown here: http://www.hitl.washington.edu/artoolkit/documentation/tutorialmulti.htm or is something wrong in my objectdata formatting etc.?

Additionally, I realised I was needlessly recreating cv.jit.threshold above, so I've switched to that for much more stable marker detection in changing lighting conditions. I'd like to find a shader that does something similar, as all the cv objects can slow things down a lot, but I've yet to come across one other than a pixelbender script.


It's supposed to run on 4 independent markers. As long as any of them track it's probably not an issue with object_data.

Which object are you using again? Can you remind me? Is it the most current one with the fixed path on PC?

Someone pointed out a bug to me recently which I've been working on but it sounded more innocuous than the one you described. Basically they all tracked but the first element of the lists (visible 1/0) was not updating properly if all objects were removed from the scene at once.

If you are using the PC version it's possible I mucked something up in the translation between Mac and PC, I was getting some errors I'd never dealt with before in VS.

If you send me more details I will do my best to investigate. Perhaps even a patch that illustrates the problem? I find recording a little video and using that for the live feed makes it easier to replicate the issue.

It's supposed to run on 4 independent markers.  As long as any of them track it's probably not an issue with object_data.

Which object are you using again?  Can you remind me?  Is it the most current one with the fixed path on PC?

Someone pointed out a bug to me recently which I've been working on but it sounded more innocuous than the one you described.  Basically they all tracked but the first element of the lists (visible 1/0) was not updating properly if all objects were removed from the scene at once.

If you send me more details I will do my best to investigate.  Perhaps even a patch that illustrates the problem? I find recording a little video and using that for the live feed makes it easier to replicate the issue.

Hey cap10subtext, thanks for the response. No worries about the inconvenience - thanks for releasing and supporting the object in the first place!

I've uploaded a quick demo video using the help patch (with added thresholder) here: 

http://www.youtube.com/watch?v=m4pBPfnWhPg

You can see it clearest by watching which message boxes get updated (only one per frame). I haven't been able to recognise a pattern to which marker it 'chooses' yet. Please forgive the noisy signal too - its night here.

I'm using the most recent object- but I've checked the previous one and it shows the same behaviour.

Here is the content of my multi directory: 

*** Initializing ARToolkit for Max/MSP***

Max External by Andrew Roth -- aroth21@yorku.ca

Opening Data File /Program Files (x86)/Cycling '74/Max 5.0/artkmulti support/multi/object_data2

jit.gl.render: building GL on window "artoolkit"...

Hey cap10subtext, thanks for the response. No worries about the inconvenience - thanks for releasing and supporting the object in the first place! 

I've uploaded a quick demo video using the help patch (with added thresholder) here:  http://www.youtube.com/watch?v=m4pBPfnWhPg

Here is the content of my multi directory: http://www.bferns.com/files/multi.zip

Jitter 1.7.0 installed
cosm version: 'Mad Professor'
*** Initializing ARToolkit for Max/MSP***

jit.gl.render: building GL on window "artoolkit"...


Hi, I have a problem with jit.artkmulti. I don't understand where is error.

My guess is that you haven't added the following line:

Applications/Max5/Cycling '74/init/max-fileformats.txt