You are viewing a plain text version of this content. The canonical link for it is here.
Posted to photark-dev@incubator.apache.org by Subash Chaturanga <su...@gmail.com> on 2011/08/12 17:36:42 UTC
Regarding interpreting the face API results in front end
Hi ,
When recognizing friends in an image, face API gives suggestions with the
evaluated confidence for all faces appears in the image.
So suppose there are 6 different people in an image and I am also appearing
in the picture . The API result can be like this.
Yadav 80% confidence
Subash 92% confidence
Luciano 66% confidence
Suho 40% confidence
Uma 35% confidence
Subash 18% confidence
So he finds some another one of my friends who may have not yet trained, and
recognizes as me in a very low possibility.
So how should I interpret the results in UI ?
option1 - display as it is (all 6 as above)
option2 - Choose the highest confidence if duplicates found.
I though if option 2 . WDYT ?
Thanks
--
Subash Chaturanga
Department of Computer Science & Engineering
University of Moratuwa
Sri Lanka
Blog - http://subashsdm.blogspot.com/
Twitter - http://twitter.com/subash89
Re: Regarding interpreting the face API results in front end
Posted by Subash Chaturanga <su...@gmail.com>.
On Sat, Aug 13, 2011 at 2:46 AM, Luciano Resende <lu...@gmail.com>wrote:
> On Fri, Aug 12, 2011 at 2:36 PM, Subash Chaturanga <su...@gmail.com>
> wrote:
> > Hi ,
> >
> > When recognizing friends in an image, face API gives suggestions with the
> > evaluated confidence for all faces appears in the image.
> > So suppose there are 6 different people in an image and I am also
> appearing
> > in the picture . The API result can be like this.
> >
> > Yadav 80% confidence
> > Subash 92% confidence
> > Luciano 66% confidence
> > Suho 40% confidence
> > Uma 35% confidence
> > Subash 18% confidence
> >
> > So he finds some another one of my friends who may have not yet trained,
> and
> > recognizes as me in a very low possibility.
> >
> > So how should I interpret the results in UI ?
> >
> > option1 - display as it is (all 6 as above)
> > option2 - Choose the highest confidence if duplicates found.
> >
> > I though if option 2 . WDYT ?
> >
>
I also thought of going with option 2.
>
> What if we go with option 2, and it's a mistake, is there a way for
> the user to fix it by providing the proper person information?
>
I didn't get you Luciano ? Did you mean "it's a mistake" referred to the
face API results with duplicates. This mainly happens when you trying to
recognize friends which you have not trained before. In such case it give
the best suggestion and thus a duplicate created.
>
>
> --
> Luciano Resende
> http://people.apache.org/~lresende
> http://twitter.com/lresende1975
> http://lresende.blogspot.com/
>
--
Subash Chaturanga
Department of Computer Science & Engineering
University of Moratuwa
Sri Lanka
Blog - http://subashsdm.blogspot.com/
Twitter - http://twitter.com/subash89
Re: Regarding interpreting the face API results in front end
Posted by Luciano Resende <lu...@gmail.com>.
On Fri, Aug 12, 2011 at 2:36 PM, Subash Chaturanga <su...@gmail.com> wrote:
> Hi ,
>
> When recognizing friends in an image, face API gives suggestions with the
> evaluated confidence for all faces appears in the image.
> So suppose there are 6 different people in an image and I am also appearing
> in the picture . The API result can be like this.
>
> Yadav 80% confidence
> Subash 92% confidence
> Luciano 66% confidence
> Suho 40% confidence
> Uma 35% confidence
> Subash 18% confidence
>
> So he finds some another one of my friends who may have not yet trained, and
> recognizes as me in a very low possibility.
>
> So how should I interpret the results in UI ?
>
> option1 - display as it is (all 6 as above)
> option2 - Choose the highest confidence if duplicates found.
>
> I though if option 2 . WDYT ?
>
What if we go with option 2, and it's a mistake, is there a way for
the user to fix it by providing the proper person information?
--
Luciano Resende
http://people.apache.org/~lresende
http://twitter.com/lresende1975
http://lresende.blogspot.com/