[Digikam-devel] face recognition in digikam & sponsoring of face recognition

[Michael G. Hansen]

kunal ghosh wrote:
>> http://websvn.kde.org/trunk/extragear/graphics/digikam/project/ImageAnnotation.odt?view=log
>  Quote from above doc
> "When designing a system for describing regions of an image, care
> should be taken to make this system extensible to other media types,
> like videos or audio files, which are already produced by today's
> digital cameras. At the same time, the system should be flexible to
> allow recognition of things other than faces to be added later, for
> example barcodes which can be used when a catalogue of items along
> with their identification numbers are photographed or scanned.
> The capability for the description of regions should match those for
> entire images, providing a caption, a description (in multiple
> languages) and tags."
> 
> Isn't it too much to ask for in one shot ?

Well, not all of this has to be implemented in one shot. But especially
the design of the data structures in the images should be flexible so
that the data structures can be extended later. And if we use XMP, that
is no problem.

>> I do not have any experience with face recognition algorithms, but since
>> you (Kunal) are working with them, it might be a good idea to test
>> whether you can find an algorithm that works well with a set of real
>> world photos and compare that with other projects out there (Picasa, for
>>
> would the real world photos be any different ? just an additional, though
> not very easy, step of face detection would be added.

By 'real world photos' I mean photos that you would actually want to
tag, like people at a party where the lighting is bad, faces are not
straight to the camera, etc. In spite of biometric photos on a passport ;-)

Michael