As many users, I think that a face detection plus a face recognition feature in F-Spot is a must-have. There has been several attempts to implement such features (firstly face detection + region tagging), but none has reached the point of being integrated in the F-Spot code (yet).
However, recently Ali Servet Donmez stepped up to implement face detection and region tagging [1]. He made some successful calls from Mono to the openCV demo code [2]. Its next steps seems to integrate such calls into the mono code, then implement region tagging inside the F-spot DDB. Personally, I can't work on such code because of a lack of time, but I would like to : 1) Raise interest of the f-spot developer/users community, to get advices, features, architecture requests, etc... (standard open-source development process) 2) Propose some algorithms ideas/libraries existing for face detection and/or face recognition. The actual opencv code (part of the demos of opencv) is not as efficient as lots of algorithms presented in recent research papers. For example, [3] proposes a solution based on AdaBoost + color detection which succesfully detects faces at about 90%, from a standard library of pictures used in research papers. When looking at the very first attempts made by Ali Servet Donmez, it seems pretty clear that the detection rate is far less than 90%. So, does anyone here knows of some open-source algorithm, based on opencv or as a library callable from Mono code, able to detect faces, or even recognize them ? Or some starting point(s) ? If you have any ideas, or advices for Ali, he always wrote "any advices welcome" on his blog ;-) and he told me he subscribed to the F-Spot ML. Thank you. Etienne Lepercq [1] : http://www.pittle.org/weblog/f-spot-face-detection-2_433.html [2] : http://www.pittle.org/weblog/face-detection-now-works-with-mono_470.html [3] : Yan-Wen Wu, Xue-Yi Ai (2008), Face detection in color images using AdaBoost algorithm based on skin color information, Knowledge Discovery and Data Mining, 2008. WKDD 2008. First International Workshop on _______________________________________________ F-spot-list mailing list [email protected] http://mail.gnome.org/mailman/listinfo/f-spot-list
