First we had Riya facial recognition (or will sometime soon, soon, really, no we mean it this time, soon, and soon (sorry)).
From a software visualization perspective the interface is pretty straight forward. I am not so sure the results are in any way meaningful, so perhaps this is more applicable for a modern art project on the west coast for a creative diaspora.
Here are my results. First the sketched image I drew – a celtic thinga-a-majig. The second page is the image I clicked on which is probably the result of some sort of ethnographic male selection process that if reviewed in great detail would somehow, someway, embarrass me.
All kidding aside, from a social software perspective Retrievr is doing it right. You can NOT and will NOT EVER get it right developing it yourself. Social software requires feedback which requires users. Pre-release your beta with plenty of "set expectations" and then listen attentively to the feedback to improve.
For example, retrievr isn’t about matching the sketch to the image. It is about matching the human THOUGHT behing the sketch to the image. Different animals. Invest in Retrievr before Riya because their developers get it.