Hi Chaps,

I’m just about to start playing around with the Kinect sensor with a view to building some vision and object recognition etc in to ALF.

I must admit, I’m a complete newbie when it comes to image processing with regard to vision and was really just wondering if any of you guys on here had any pointers/tips on anything Kinect related (if you have used it) or any good, decent vision links (papers, API’s etc).

OpenCV seems to be the defacto, so I’ll be starting with that.

Just looking for thoughts really

Dan

Hey, Dan, have you looked into the Microsoft Kinect SDK yet? I realize that the SDK is for use on a Windows platform (ALF is on a Linux-based cluster, is it not?), but since the SDK has resources for working in C#/C++, it shouldn’t be too difficult to write executables for Linux (if you have a Windows machine, that is)

I’ve also read quite a few articles at robot.net that involve using Kinect. You may find some useful resources there.

This video has a flying robot ... (and other hacks)

http://www.youtube.com/v/ho8KVOe_y08

Sounds like you are on the right track.

Please go on about it ...

Hi Dan,

Sounds like a good idea. There are not many chatbots around responding to other input than typed text. Would you love to hear more about your progress!!!

Interesting area of work. Could the $100,000 Loebner gold medal be at stake after a bit more development by Dan?

heheh, I wouldnt go that far!  But it’d be a nice hope that some of what Im working on at the moment can maybe push the frontier a bit more

Unfortunatley over the past 2 weeks or so Ive had some other matters to attend with my business, lots of meetings and jetting about….which ultimatley results in lots of wasted time hanging about in airports, so not as much has happened that I would of hoped.

However, I did start to play about with the Kinect and some vision idea’s before business commitments were thrust upon me and I did make a bit of initial progress.

So far I am able to detect objects with the Kinect, track them and extract information about them (I believe they are reffered to as blobs within the vision communities).  Ultimatley after playing with the OpenCV libraries for a couple of days, I decided it wasn’t for me.  While it is a very powerful library, it didnt feel like it “jelled” with the architecture I was going for and getting it to interface with some of the early system I had already put together was proving difficult.

Instead I have opted to go for a more fragmented approach, which is more flexible, thus I am now running separate Kinect drivers, along with the specific image processing routines I require for my method.  I’ve had much more success with this approach.

Hopefully during the course of the next week or so I should be able to take my usual back seat in the day to day running of the business and get back into this.  I’m aiming to start looking at actual object recognition now that I have the tracking and detection working to some degree.

Until later

Stay well and best regards

Dan

Wouldn’t it be nice if a chatbot could detect a users emotion and react accordingly…. You look unhappy today….. I see that answer made you smile…

I guess that could be done with a webcam link rather than needing a kinect attached.

The kinect interests me as well (beyond playing kinectimals and dance central with my daughters) but I have doubts as to its accuracy with regards to providing use-able mocap data for animations.

In the mean time, we’ve added companies in biometrics:
http://www.chatbots.org/companies/biometrics
and
facial recognition as well:
http://www.chatbots.org/companies/other/#face_recognition

Our next step is to identity their consumer-ready products. (we’ll love help btw to test all these stuff!)