Re: Core Image Color Tracking in GPU Gems 3

Subject: Re: Core Image Color Tracking in GPU Gems 3

From: Robert Douglas <email@hidden>

Date: Thu, 27 Mar 2008 11:20:59 -0700

Delivered-to: email@hidden

Hi Ian,

Did you get this resolved? I haven't bought (or seen) the book, but I saw a demo at WWDC and I presume the code is similar to what is in the book. Apple hasn't published it as a sample program yet so I am still bound by the NDA and I can't just send you the source. And as I don't have the book...

I've run CIColorTracking on several machines without any problem, including my MacBook, so it should work fine on your MBP. And yes the duck should follow the ball around.

I also tried implementing similar algorithms in Quartz-Composer by writing my own kernels, but was not happy with the performance. The Area Average filter is a bottleneck, and as I need to track multiple points, I've gone back to doing everything on the CPU. Is the Color Tracking Area Mean Filter version any better?

Cheers,

Rob

---

Dr. Robert M. Douglas,

Department of Ophthalmology and Visual Science,

University of British Columbia,

email@hidden email@hidden

http://www.cmr.ubc.ca/douglas.html

http://www.robertmdouglas.com

On 26-Mar-08, at 11:17 AM, Ian Grant wrote:

Dear all,

Sorry to bother you all with this one. It may not be completely on topic:

I received GPU Gems 3 yesterday and tried out the CIColorTracking example.

The nib in the sample code in GPU Gems 3 appears to be zero bytes.
Does anyone know of an errata page for GPU Gems 3 - I searched but couldn't find?

I quickly constructed a new nib but seem to have an issue. I assume the duck should follow the pink ball ;-).

Image: New Color Tracker Nib

<http://ellington.tvu.ac.uk/images/CIColorTracking001.png>

I am building on Leopard (10.5.2) on a mac book pro. I assume the code is not nvidia specific. I even downloaded xcode 2.5 to see if there were simpler issues with the NIB.

Also: I attempted to use the CI filter(s) in quartz composer, but couldn't quite work out the logic. I guess I could be witnessing the same problem exhibited in the application.

Image: Quartz Composer Patch using CI Color Tracker units

<http://ellington.tvu.ac.uk/images/QuartzComposer001.png>

The 'Color Tracking Area Mean Filter' outputs an image of bounds: {{0,0},{1,1}} pixels.
I've also experimented by setting each CI filter to 'Use Normalised Coordinates'.

Again, any hints or tips would be welcome.

Kind regards, Ian

*******************************
Ian Grant
Senior Lecturer in Digital Art
Faculty of the Arts
Thames Valley University
Ealing, UK
W5 5DX
<email@hidden>
*******************************
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Quartz-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/quartz-dev/email@hidden

This email sent to email@hidden

_______________________________________________ Do not post admin requests to the list. They will be ignored. Quartz-dev mailing list (email@hidden) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/quartz-dev/email@hidden This email sent to email@hidden

References:
	>Core Image Color Tracking in GPU Gems 3 (From: Ian Grant <email@hidden>)