[sldev] Anyone here with OpenCV experience?

[Moriz Gupte]

And I think the Wii functioning perfectly out the box proved that such
technologies may certainly become mainstream once variations due to
environmental conditions are dealt with properly. I think calibration is the
other thing that needs to be automatic (there are strategies for doing
this), a possible algo for e.g. to automate eye tracker calibrations could
use the info we know about reflex gaze behavior with respect to animated
visual stimulus (this is an approach I once implemented because austistic
kids were not expected to follow instructions before their gaze was
tracked). Well it does not 'stictly eliminate' the calibration but it
happens without the user noticing it and it can be pretty fast (10-15s  for
a single pass ...) next question is...the advantages for tracking movement
needs to be really good to outweigh the various Usability issues that will
creep in (does not mean we should not try though)

On Thu, May 21, 2009 at 4:57 PM, Jan Ciger <jan.ciger at gmail.com> wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Dahlia Trimble wrote:
> > Overkill? I'm not so sure. Consider the extreme variations in contrast
> > that something like this would be exposed to out in the field. Lighting
> > is uncontrolled and unpredictable. A major source of light on a user's
> > face may well come from the image shown on the monitor, and may reflect
> > off of the user's glasses. Facial features are variable and skin tone,
> > eyebrow color, and other facial hair could further confuse any detection
> > methods. Background lighting and objects other than the user could cause
> > false detections. Also the camera could not be controlled, Could one
> > reasonably expect users to purchase a standard webcam from a single
> > vendor when many computers come with integrated webcams?
>
> No, but that isn't the point. It doesn't need to be 100% perfect and
> robust - even the Logitech's solution is not. And yes, actually most of
> the lighting in an average room is stable - and can be compensated for.
> For example, the paper I have sent the link to deals with the lighting
> issues by calibration - it takes a picture of a skin patch (your face)
> and uses that to create a model of your skin color. That compensates for
> all the skin and lighting issues, even lot of camera parameters.
>
> An actual application would probably build a background model too to
> make the detection more robust. Again, a rather simple task.
>
> >
> > Somehow I don't think simple methods demonstrated that in a lab
> > environment will perform well at all given all these potential confounds
> > in all user environments. Whatever is implemented would need to be
> > subjected to extensive user testing if any kind of wide acceptance would
> > be desired.
>
> Sony is doing a brisk business with their hugely popular EyeToy series
> that is using these simple algorithms that you believe do not work. Do I
> need to say more?
>
> Regards,
>
> Jan
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.9 (GNU/Linux)
> Comment: Using GnuPG with Mandriva - http://enigmail.mozdev.org
>
> iD8DBQFKFdxAn11XseNj94gRAioWAJ4xNh7wHz5RZiYvGmtKupkuIh+bBQCdGHih
> wvsn1lnJPRQPNBn+LCOe+iw=
> =72oW
> -----END PGP SIGNATURE-----
> _______________________________________________
> Policies and (un)subscribe information available here:
> http://wiki.secondlife.com/wiki/SLDev
> Please read the policies before posting to keep unmoderated posting
> privileges
>



-- 
Rameshsharma Ramloll PhD Research Assistant Professor Idaho State
University, PocatelloTel: 208-282-5333
More info at http://tr.im/RRamloll
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.secondlife.com/pipermail/sldev/attachments/20090521/9c29243e/attachment.htm