Orbeus’s visual recognition engine and its network implications
Computers are improving in their ability to recognize items in normal photographs. A field called computer vision utilizes advanced algorithms to make inferences about data in pictures. You may have already seen examples of image recognition when your camera recognizes the face(s) in a scene before you shoot a picture. Or perhaps you have noticed Facebook suggests who a particular person might be based on his/her facial features. One company—Orbeus—aims to to make computer vision even smarter; it plans to empower software with algorithms that go beyond simple facial recognition.
Not only is Orbeus’s software capable of realizing gender, estimated age, and emotions, but it also possesses the ability to recognize logos, text, and other non-human objects. It can even identify the type of scene where a photo was taken (e.g., a forest, a beach). It does so, in part, by comparing the image to photos of known objects within its database. Using a network-like process (how strong are the edges between a cluster of nodes depicting known objects and the image of the object in question?), the software creates a best-guess for what the image might be.
The software has numerous potential applications (companies are already making use of its API), but perhaps its most powerful lie in the realm of social networking and targeted advertising. For instance, currently, social networks typically rely on user-inputted data (favorite music, movies, activities) to compute the strength of a tie between two individuals. However, with Orbeus’s software, this could be done automatically. The photos of two individuals might be compared to determine how many they appear in together, whether they were taken in similar locations, or whether they depict similar activities. Utilizing this data, the strength of the edge between these two people can be calculated automatically through photographs, possibly presenting an opportunity to make better friendship recommendations.
This ability will benefit both social network users and companies that display their ads on these websites. Advertisers will be able to target clusters of individuals that would more likely buy their products (i.e., if a woman has multiple images of her playing tennis using a racket produced by a certain company, they might want to show her an ad for their latest racket model). They could also use data to determine the strength of the edges between certain demographics and their product (i.e., are there more men or women in photos with a certain product?) and tailor their advertising techniques to better fit their findings.
Orbeus’s software presents powerful implications for the future of social networking, allowing for the automatic creation of such networks based on images alone.
Sources:
http://www.fastcompany.com/3000905/orbeus-trains-machines-recognize-faces-emotions-broccoli
http://orbe.us/
– Brian