We use a combination of computer vision and human crowdsourcing to always give results to any visual query (an “any image” platform). Ideally, when an image query is received, the key visual content is identified with our computer vision modules within a few seconds, and that label is returned via the API. If we can not get an accurate result in computer vision, it goes to a suite of crowdsourcing nodes, an innovative network architecture that involves tagging, gaming and social interaction with photos. For IQ Engines, the crowdsourcing technologies are important because they train (a feedback loop) the computer vision modules. This is what we call “real time learning”, where with every query submitted the system learns. You can see a demonstration from last year: http://vimeo.com/14310752. Over the next year, the system will improve as we add more CV modules (we are actively working on 3D object building and searching), and innovations in expert crowds. We built this “Vision as a Service” platform via grants from the National Science Foundation and a strong team of experts in computer vision. It’s being used by several large companies to power visual search applications. The system works best if you first train the CV with labeled object images – we recently released a Training API so you can add objects to our database.