Installing OpenCog on Raspberry Pi 3, Raspbian Jessie

[Dave Xanatos]

I wish to get OpenCog and all dependencies/bindings running on my Raspberry Pi 3, running Raspbian Jessie.

I also currently have running OpenCV, CMUSphinx/PocketSphinx, ESpeak, and many supports (SciPy, NumPy, SKLearn, etc.....) - these are all working well together, primarily running face detection/recognition, eye tracking via servos connected to the "eyes", etc.

I do not wish to clobber my current build or any part thereof, and wish to be sure I am using the correct installation procedure.  I see octool is available for installation on Ubuntu systems, but there seem to be dependency differences between Ubuntu and Debian/Raspbian.  If I am correct, the dependencies must be installed correctly before OpenCog can be installed.

The Debian-specific dependencies installer script I see links for all go to a 404 -page not found.

Despite the fact that my current build with OpenCV et al might suggest I have some idea of what I am doing, please don't be fooled! :)  I am just very good at following clear instructions and pick up new programming languages quickly (Python being the latest).  Two months ago I didn't even have any Linux experience and knew nothing of Python, but once I get my "foot in the door", I gain altitude very rapidly.

I'm hoping to hear from anyone who has successfully installed OpenCog on a Raspberry Pi 3 running Raspbian Jessie, with Python3.  

Thanks for your help, I hope to be able to contribute something useful here.

Dave

[Roman Treutlein]

While this sound like a cool project. Why exactly do you want OpenCog on your Pi? OpenCog is quite resource intensive i can't think of any use case where you would want to run it on a pi. Somehow connected and let it control a pi maybe but not run on it.

That said you might have some issues with Python 3 because OpenCog still only supports Pytoon 2.7.

[Dave Xanatos]

The ultimate goal will be a distributed system - a three-Pi system, specifically.  I chose R Pi for a variety of reasons, and so far I have been very impressed with its performance handling such tasks as openCV was designed for.  The Pi doesn't seem to be even remotely resource-starved in running all the things I have going on with it now.

The initial development system is getting me up to speed, and allowing everything to play together in a single box.  Ultimately, there will be a separate Pi for the left "eye" and left "ear", a near twin for the right eye/ear, and a third for OpenCog and speech output.  "Unconscious" functions, such as eye tracking with servos, image preprocessing, movement detection, sound detection and location assessment, and image/face identification, will be handled by the left and right Pi, with intercommunication between both halves and the OpenCog Pi.  "Conscious" functions will be handled by the third Pi, the OpenCog Pi.  By Conscious functions, I specifically mean those functions for which decisions and determinations must be made that can only be made by the OpenCog architecture.  I think of it an analogous to left-hemisphere, right-hemisphere, and reptilian brain-stem

I have also been told that the Pi can run everything but MOSES, which maxes out RAM, but by adjusting gcc it can probably also be gotten to run.  For my initial build, I really only need to ensure I figure out correct installation and dependency setup, and that I can get interoperability between my other programs and OpenCog - when I find my system actually straining for resources then I will level-up to the three-Pi build

As for the differences between Python 2.7 and Python 3, they are minor and my current build has provisions for both, I do not anticipate that to be an insurmountable issue.  I chose primary development to go forward under Python3 because, as they say, "that's the future of Python" :)

Dave