Re: Feature request

[Matthew Brush <mbrush codebrainz ca>]

On 01/31/2012 08:11 AM, Antono Vasiljev wrote:
> On 01/31/2012 12:51 AM, Stefan Sauer wrote:
>
> > > > > Any advise?
> > > > You could use CheeseAvatarChooser, which presents a dialog to take a
> > > > photo, and then provides a frame for cropping the result:
> > > >
> > > > http://developer.gnome.org/cheese/unstable/CheeseAvatarChooser.html
> > > >
> > > > This is the widget that is used in Empathy and the User accounts panel
> > > > in the Control Center. The API is simple, and gives a GdkPixbuf of the
> > > > resulting image.
> > > Hmm. Not sure it provides needed functionality.
> > >
> > > Time lapse video will be smooth if only user face on each photo
> > > positioned the same each day. It means user should make photo from same
> > > distance (fit her face to some visual box on the screen) and use some
> > > visual markers for face centering. No need to crop photos if they made
> > > with same configuration each time.
> > >
> > > So question is: how hard it's to position some SVG image (calibration
> > > box) over camera view?
> > The underlying gstreamer plugins have most features you will need. E.g.
> > facedetect can now also detect mouth, nose and eyes. You could also show
> > the previous picture using videomixer with opacity to support positioning.
>
> Sounds interesting. Got it.
>
> But if this feature request should be implemented outside cheese (as
> someone pointed earlier) then i need something similar to android
> intents here. I mean this flow:
>
> - app requests photo from camera
> - cheese appears and ready to make photo
> - when photo made it returned to the app
> - cheese closed, my app happy dealing with photo
>
> Is it possible somehow or only with avatar-chooser widget?

You don't really need Cheese or special stuff at all for this, you can 
just use GStreamer and/or OpenCV directly in a number of different 
programming languages and operating systems. In fact, you might find it 
harder to make these other programs bend to your will rather than just 
use what they use for video/pictures in the first place.

Just make a window with the live video feed in it (lots of examples on 
the web) and a button to take the picture. You can use GStreamer, Cairo, 
OpenCV or other stuff to put an overlay, or as was suggested you can use 
OpenCV face detection[1] to line up the faces, since it's able to tell 
you about how big a face is, where it's at, and apparently even features 
like the mouth.

IMO, you should first write a small program[1] to show live video from 
camera. Once you do that, the rest will become far more clear.

Cheers,
Matthew Brush - cheese-list lurker

[1] I wrote this some time ago as a simple example of face detection, 
you could start with this and go from there:
https://gist.github.com/1441735