Photo sharing site Riya.com has pioneered new technologies in graphical search engine technology. Using highly advanced facial recognition, they are able to auto detect faces from within your online photo album.
It works through a training program. This is not unlike other similar technologies such as speech recognition. It requires several pictures of a person to be “identified” and then it can begin to “find” the person in other pictures.
Overall, the algorithms involved in such a search can be highly complex. This is a much higher complexity than fingerprint analysis. In fingerprint analysis, a program can map points where ridges split or join, and then calculate a distance and direction between each. When recognizing a fingerprint match, it simply checks for a proportional picture of these points. If it is basically the same, the software identifies a match. If not, the print is rejected.
Facial recognition is vastly more complicated. For more details on facial recognition algorithms, check out face-rec.org’s page on facial recognition algorithms. There are several algorithms that can be used. An algorithm that I thought sounded “easier” was the Elastic Bunch Graph Matching Algorithm. It uses a graph and vectors to map the face based on fiducial points (such as noses, eyes, etc.).
By the way, Riya.com is currently giving away unlimited space to store your photos and has an automatic upload tool for your My Pictures folder on your computer. It is worth your time just for that. I found the web interface easy and intuitive, and overall, I am quite satisfied with the site.