I have been following Riya – a photo sharing and searching web app for quite some time. It is interesting to see that someone is trying facial recognition at this scale. Facial recognition is a really hard problem to solve. In a nut shell think of a complex graph formed by common facial features like – distance between eyes, width of the nose etc to form a faceprint. Then these nodal points are compared to a reference set of images in a database.
Easier said than done of course!
When one talks about photo sharing sites, Flickr cannot be left out of the conversation. I was thinking about how fundamentally the two approaches differ. What it boils down to is – the former using computing power and the later using the social computing power of the web to make images searchable.
So I wonder which is more accurate and what kind of results would one produce versus another?
One thing is for sure; computers can recognize but not interpret.