Re: [Rtftr-devel] training accuracy

[Abhishek D. <adu...@gm...>]

Hi Chris,
Yes, that's the easiest way to improve accuracy but this entails high
computational cost during the face comparison process.

Another way to improve accuracy for the kepenekci module would be to fine
tune the frequency and orientation of Gabor filters being used to compute
features. Furthermore, you can devise some machine learning algorithm to
model when the subspace LDA and Kepenekci algorithms fail and use this
model to weight the decision from these two modules: a type of decision
fusion.

But again, face recognition is a difficult problem it is extremely
difficult to find a strategy or algorithm that succeeds in every case.

Good luck
Abhishek

On Mon, Apr 23, 2012 at 5:19 PM, Chris Hellberg <ch...@ch...>wrote:

> Hi,
>
> I'm running the app and it manages to recognise faces but it seems to miss
> several of them. Is the key to increasing the accuracy of recognition to
> seed as many diverse 64x64 photos into the data/kepenekci/trainfaces data?
>
> Cheers,
>
> Chris
>
>
> ------------------------------------------------------------------------------
> For Developers, A Lot Can Happen In A Second.
> Boundary is the first to Know...and Tell You.
> Monitor Your Applications in Ultra-Fine Resolution. Try it FREE!
> http://p.sf.net/sfu/Boundary-d2dvs2
>
> _______________________________________________
> Rtftr-devel mailing list
> Rtf...@li...
> https://lists.sourceforge.net/lists/listinfo/rtftr-devel
>
>


-- 
Abhishek
http://abhishekdutta.org