[Fiware-miwi] Hands-on: Augmented Reality GE

Antti Karhu antti.karhu at cie.fi
Fri Feb 21 10:08:28 CET 2014


Hi,

Good questions, hope this answer will clear things

   - There is the crucial need for markerless tracking using predefined
   image markers in FIcontent. It is one of the features of the utilized ALVAR
   library. When will it be exposed through JavaScript and hopefully also
   Xflow?

Let's start with some definitions:

(1) Marker: A fiducial marker is defined as a square box that is divided
into a grid. The outside cells of the grid are always black and the inside
cells can be either white or black.

(2) Image Marker: An image marker is very similar to the Marker with the
difference being that the resolution of the grid is a bit larger and the
inside cells can form a shape or logo.

(3) Natural Feature Image: A natural feature image is simply an image, it
might be computer generated or taken by a camera. The image doesn't have
any borders nor doesn't necessarily have to be square or rectangular.

There are multiple versions of ALVAR with different level of features
supported. The one that have been ported to javascript is called ALVAR
MOBILE. It supports detecting and tracking markers(1) and  image
markers(2). The image markers must have a clearly recognizable border and
the contents must not have too much detail but rather be made up of solid
shapes or big block letters. The ideal resolution for image markers is
spanning from 8x8 to 32x32.

There is already a version that exposes the detecting and tracking image
markers through Xflow:
https://github.com/Cyberlightning/Cyber-WeX/blob/master/DataflowProcessing/demos/xflowAR/index.php.
I haven't included this to the AR GE yet, because there is serious
problems with accuracy and performance with mobile devices.

There is also an another version called ALVAR Tracker, that implements a
generic 6DOF tracker but can also be adapted to detect and track (3)
natural feature images. This version is not ported to javascript nor won't
be ported in near future.

The ALVAR Mobile was ported to javascript using emscripten and this was
quite an easy process because the ALVAR Mobile doesn't have any third party
dependencies. However this isn't the case with the ALVAR Tracker. The guys
on VTT thought that trying to port it to javascript is much bigger task,
and because it has a multi-threaded AR pipeline with hardware
acceleration(SIMD, NEON) that can't be currently achieved with javascript,
it is basically just  waste of time.


   - To what degree do you take advantage of hardware acceleration, in
   particular to be used on mobile devices?

There is no hardware acceleration used now, and no plans to use it for next
release 3.3. The situation is that the emscripten ported ALVAR is machine
generated javascript, so it's pretty much a black box, you put the video
frame in and get the transform matrix out. Trying to identify which part of
the code could benefit from hardware acceleration is hard. Even harder
would be to replace parts of the javascript code with an equivalent GLSL or
WebCL parts. Obviously someone have to write the shaders of those parts
first.  Because of the limited resources, unfortunately there is no change
that this will happen at any time soon.


 -Antti-


2014-02-20 20:58 GMT+02:00 Stefan Lemme <stefan.lemme at dfki.de>:

>
> Hello,
>
> after our workshop last week a few questions about the Augmented Reality
> GE raised and hopefully you can provide some insights.
>
>    - There is the crucial need for markerless tracking using predefined
>    image markers in FIcontent. It is one of the features of the utilized ALVAR
>    library. When will it be exposed through JavaScript and hopefully also
>    Xflow?
>    - To what degree do you take advantage of hardware acceleration, in
>    particular to be used on mobile devices?
>
> I would really appreciate to get some more informations about the open
> questions.
>
> Best regards,
> Stefan
>
> --
> ********************************************************
> Stefan Lemme
>
> DFKI GmbH
> Agenten und Simulierte Realität
> Campus, Geb. D 3 4, Raum 0.75
> 66123 Saarbrücken
>
> Tel.: +49 (0) 681 / 85775 - 5391
> Fax: +49 (0) 681 / 85775 - 2235http://www.dfki.de/web
> ********************************************************
> Deutsches Forschungszentrum für Künstliche Intelligenz GmbH
> Trippstadter Straße 122
> D-67663 Kaiserslautern, Germany
>
> Geschaeftsführung:
> Prof. Dr. Dr. h.c. mult. Wolfgang Wahlster (Vorsitzender)
> Dr. Walter Olthoff
> Vorsitzender des Aufsichtsrats:
> Prof. Dr. h.c. Hans A. Aukes
> Sitz der Gesellschaft: Kaiserslautern (HRB 2313)
> USt-Id.Nr.: DE 148646973
> Steuernummer:  19/673/0060/
> ********************************************************
>
>
>
>
>
> ********************************************************
> Stefan Lemme
>
> DFKI GmbH
> Agents and Simulated Reality
> Campus, Build. D 3 4, room 0.75
> D-66123 Saarbruecken
> Germany
>
> Phone: +49 (0) 681 / 85775 - 5391
> Fax:      +49 (0) 681 / 85775 - 2235 http://www.dfki.de/web
> ********************************************************
> German Research Center for Artificial Intelligence
> Deutsches Forschungszentrum fuer Kuenstliche Intelligenz GmbH
> Trippstadter Strasse 122, D-67663 Kaiserslautern, Germany
>
> Management Board:
> Prof. Dr. Dr. h.c. mult. Wolfgang Wahlster (Chairman)
> Dr. Walter Olthoff
>
> Chairman of the Supervisory Board:
> Prof. Dr. h.c. Hans A. Aukes
>
> Amtsgericht Kaiserslautern, HRB 2313
> ********************************************************
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.fiware.org/private/fiware-miwi/attachments/20140221/bcc78bdf/attachment.html>


More information about the Fiware-miwi mailing list

You can get more information about our cookies and privacy policies clicking on the following links: Privacy policy   Cookies policy