libface integration

[Aditya Bhatt]

Hi Everyone!

@Marcel
Hi, you can checkout the code for libface from this svn repo:
 
svn co https://libface.svn.sourceforge.net/svnroot/libface libface

As of now, only jpg images are supported, and supporting other common formats is a very simple matter involving only a handful of lines of code.
Since we tried to be API-agnostic in libface, the only way to load an image for detection was to load it from the filesystem. So we'll have to make a variant of libface exclusively for digiKam, but it shouldn't be much work, only changing one or two functions. But I can do this only after I get familiar with digiKam's API, which I can start after about 5 days (I have exams).

Apart from the above quirks, I guess we have all we need to integrate this into digiKam. Even after integration, I and Alex can continue working on the plugin separately, of course.

As for the image tagging widget, there is one in nepomuk playground repo, but it has build errors and seems to be abandoned. I talked with someone named "urkud" on nepomuk's IRC, and It seems that he is the developer. He himself said I shouldn't use that widget, and added that he would be committing fresh code for peopletag in his gitorious account soon : http://gitorious.org/~urkud1

I haven't made many custom widgets, but someone must have made such a thing so I'll google for now :)

And about the face fingerprint database : that'll be stored in a different file, as it will be huge in size :)


--
Aditya Bhatt
My Blog : http://adityabhatt.wordpress.com

_______________________________________________
Digikam-devel mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-devel

[Marcel Wiesweg]

Am Sonntag, 21. März 2010 schrieb Aditya Bhatt:
> Hi Everyone!
>
> @Marcel
> Hi, you can checkout the code for libface from this svn repo:
>
> svn co https://libface.svn.sourceforge.net/svnroot/libface libface
>
> As of now, only jpg images are supported, and supporting other common
> formats is a very simple matter involving only a handful of lines of code.

Don't implement your own image loader - accept loaded image data.

What is the configuration data for which you accept a directory? Which options
are set there, and from whom? Or is this where the training data is stored?

I also see a hardcoded path to opencv installation, which is something which
will break in real-life...

> Since we tried to be API-agnostic in libface,

We will want a nice Qt-based layer, but I dont see any problems right now why
this shouldn't be possible on-top.

> the only way to load an image
> for detection was to load it from the filesystem. So we'll have to make a
> variant of libface exclusively for digiKam, but it shouldn't be much work,
> only changing one or two functions. But I can do this only after I get
> familiar with digiKam's API, which I can start after about 5 days (I have
> exams).

We dont have an API in a strict sense ;-) We are rather built on a platform,
which is Qt/KDE.

>
> Apart from the above quirks, I guess we have all we need to integrate this
> into digiKam. Even after integration, I and Alex can continue working on
> the plugin separately, of course.
>
> As for the image tagging widget, there is one in nepomuk playground repo,
> but it has build errors and seems to be abandoned. I talked with someone
> named "urkud" on nepomuk's IRC, and It seems that he is the developer. He
> himself said I shouldn't use that widget, and added that he would be
> committing fresh code for peopletag in his gitorious account soon :
> http://gitorious.org/~urkud1
>
> I haven't made many custom widgets, but someone must have made such a thing
> so I'll google for now :)
>
> And about the face fingerprint database : that'll be stored in a different
> file, as it will be huge in size :)

_______________________________________________
Digikam-devel mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-devel