Discovering User Connections Using Their Shared Images on Social Media

Many social media applications, such as recommendation,virality prediction and marketing, make use of socialgraphs, which may not be explicitly specified or kept privately.Meanwhile, advanced mobile devices have become part of our liveand a huge amount of user shared images is being generated everyday. Obtaining user connection is hence possible by analyzingthose images, in which one of the most common approaches isusing user annotated tags which are not always reliable. Thisvideo demonstrates how non-user generated annotation with nonusergenerated labels helps the connection discovery by enhancingthe quality of the annotation with computer vision techniques.Users are represented by the occurrences of each unique nonusergenerated labels and connections are then discovered fromthese occurrences. The discovered connections are useful for informationor service recommendations in any social network withimage sharing, are other interesting personalization applications,even without access to social graphs.