OF on BeagleBone


After fighting with libusb I got the Kinect to work!

This video is just running libfreenect’s tiltdemo but I just tested my ofxKinect app and it is now running



jvcleave that’s awesome! I’ve been looking at trying to get my Pandaboard to work with my Kinect, but really haven’t had much time to play with it yet. I have OF more or less up and running, but I haven’t really dug into interfacing any camera with it yet. Any thoughts on how you might do graphics output?


Thanks Joshua, I was thinking of using it to measure distance changes being headless.

OpenNI is next for me, I had it compiled previously but the libusb problems still plagued me from connecting to the camera.

I have seen these demos of graphics with DVI addons (capes)


And this one looks awesome (but out of stock) http://tigal.com/product/2410


another cheap way would be to maybe go web based with a nodejs module?


jvcleave this sounds suitable:



How big was that SD Card that you installed Debian on and as a secondary q, why is it so huge?

base@factoryfactory ~/Downloads $ zcat beaglebone_debian_of_cloud9_no_FIBERS.img.gz | sudo dd of=/dev/mmcblk1p1 bs=64M
dd: writing `/dev/mmcblk1p1’: No space left on device
0+117255 records in
0+117254 records out
3903848448 bytes (3.9 GB) copied, 1469.91 s, 2.7 MB/s

Did I do something wrong or does it just require a 8GB microSD?


I am pretty sure I have it on some 4GB cards but I did buy some 8GB ones later. I was running out of space at one point as I was pulling OF from github. There may be a lot of build files as I was also compiling a bunch of NFC stuff, building Cloud9/node from scratch…


Actually, I think this crapped my sd card, as I see this when I stat it now:

stat /dev/mmcblk1
File: `/dev/mmcblk1’
Size: 0 Blocks: 0 IO Block: 4096 block special file
Device: 5h/5d Inode: 23308 Links: 1 Device type: b3,8
Access: (0660/brw-rw----) Uid: ( 0/ root) Gid: ( 25/ floppy)
Access: 2012-03-07 07:56:10.156049959 -0800
Modify: 2012-03-07 07:56:10.156049959 -0800
Change: 2012-03-07 07:56:10.156049959 -0800

Oops. GParted won’t let me partition it and errs out trying to erase it. Ouch.


That sucks :frowning:

@xach (hope you are on notify) - what size card did you use?


Hi Jvcleave, i m trying to use OF within a Pandaboard. Actually i’ve installed Ubuntu server 11.10 on my PandaboardES, following these instructions https://sites.google.com/site/andrewssobral/pandaboard ; but i i had a lot of troubles . I have used Damianz armv7 version of Openframework from Arturo github repo. I had issue with OpenGLES because i had omap sgx opengles lib installed instead of libgles-mesa. Running the script install_dependencies.sh was succesful. But when i create a new codeblocks project with the python script (that in path_to_openframeworks007/scripts/linux) and compiled the project, i receive error about vfp not matching . maybe it’is not a good idea run the python script for create projects…? And maybe the linux distro that you used(this: http://elinux.org/BeagleBoardDebian#SGX-Video-Acceleration ) is also suited for the PandaBoardES? this shuold helpme a lot. Maybe someone else has working with a Pandaboard and can give me some more hints…



Actually the one I used was this one (but for squeeze I believe)

I wonder if the libraries provided in the linuxarm7 branch are compiled using these flags
-mfloat-abi=hard -mfpu=vfp

that don’t match up with your system?


thanks for the info.

I wonder if the libraries provided in the linuxarm7 branch are compiled using these flags
-mfloat-abi=hard -mfpu=vfp

yes linuxarm7 use this flags . looking at elinux link i can see:

Port Details: http://wiki.debian.org/ArmHardFloatPort
Status: [http://wiki.debian.org/ArmHardFloatTodo


maybe i have to use this .


yes my branch will only work with hard float and i didn’t make opengl work yet. i know damian has managed to run it and that he is using armel (not hard float) probably it’s in his repo but not sure. also jason, didn’t you compiled it for armel for the chuby?


With the Chumby (arm5) I mostly used the Linux version with my own flags in the makefile. I pretty much removed the libraries that came with OF and used ones installed on the system from apt-get

I compiled kiss and tess2 with the make/config files you have in your branch

Cairo I had to compile from scratch as apt didn’t have the latest


The SGX libraries for Beaglebone/Panda are not built with hardfp option.

Is this a requirement for getting started with OF (noob question) ?


@prabindh not OF in general but if you are using the damian_linuxarm7 branch for an arm device then yes


@kalwalt, there are instructions for pandaboard on my damian_linuxarm7_softfp branch here https://github.com/damiannz/openFrameworks/tree/damian-linuxarm7-softfp

At the moment the sgx graphics binaries are soft float only which means you must be running a soft float operating system. Without the sgx binaries the graphics will not be accelerated and that means the performance will be terrible.


@xach copying usr/include/GL to GLES will not have solved any of your problems… GLES is a different system to GL and as such it has its own set of header files. If you’re missing a GLES include folder that means that GLES is not installed on your system. You’ll need to find the correct GLES/EGL binaries, I guess from SGX, and install those…

The complicated part is that there’s also a kernel module. In general getting 3d acceleration to work is the most difficult part of running oF on an ARM machine. The only time I’ve had 100% success was with a Pandaboard and Ubuntu 11.10, as described in the README on my damian_linuxarm7_softfp branch.


yes i follwed your instrcuctions and this works even in server edition . At this moment i can’t works on pandaboard ( due tomy works engagement in the mountains ( i work on a refuge in Alpi giulie) but will doi in the next months.
I had also problem with 3d addons (3dmodeloader and assimp) because not based on GLES . i found that someone had already used them with ARm. I’m wondering if you have experience( as i suppose in the previous post) with this and if you have some code somewhere.


Ah it seems I found this a little late. However, I was wondering if the Beaglebone would be sufficient to perform point cloud and path finding calculations with ofxKinect. Would it be preferable to have an FPGA dedicated to those calculations in addition to the Beaglebone? (Noob question)



I started something similar here.