AI Learns To Predict Human Behavior From Videos

 

As humans we are able to pick up on the many versions of body language and process whether it is good or bad, and quickly come up with a reaction to it. Some of the body language I'm speaking of a proper hello, handshake or even a wave. We may not know right away what another human will use but because of our quick processing and awareness of the surrounding area, we are able to read the situation and respond appropriately. Recently, researchers at Columbia Engineering have brought to light a computer vision technique that allows machines a intuitive sense for projecting next moves by capitalizing higher-level correlations between people, animals, and objects. According to Carl Vondrick, Assistant Professor of Computer Science at Columbia, "Our Algorithm is a step toward machines being able to make predictions about human behavior, and thus better coordinate their actions with ours". He further stated, "Our results open a number of possibilities for human-robot collaboration, autonomous vehicles, and assistive technology". The researchers claimed this to be the most accurate method that will predict live video events up to seven minutes in the future. In their study, they analyzed thousands of videos and one in particular like The Office. The system processed and predicted various activities that ranged from handshaking to fist bumping. In the past, predictive machine learning ran by other teams have concentrated on prediction of single actions at a time. This would then categorize the action as a hug, high five, handshake, or a non-action. Columbia Engineering with this knowledge chose to approach this by looking at the longer-range prediction problem from a different viewpoint. At the end, the team concluded that the new algorithm would refine accuracy of future predictions on significant tasks than in previous times. However, the next step to ensure this success is to verify that it will work in real world application and not just the lab. If the system can handle various settings, then we can be assured many opportunities await to deploy machines and robots that will greatly improve our safety, health, and security.