Researchers from University of Waterloo develop machine learning video-analysis software capable of detecting when drivers are distracted.
We’ve recently covered an initiative that gave concerned citizens a platform for uploading videos of drivers who text, as well as a smart bike lock that blocks cyclists’ calls in transit, and now researchers in Canada have developed an in-car system that aims to save distractable drivers.
The team, from the Centre for Pattern Analysis and Machine Intelligence (CPAMI), University of Waterloo, Ontario, Canada, have taken an AI approach to ‘teaching’ video analysis software how to detect when drivers are showing signs of distraction. CPAMI has previously used machine learning techniques to identify when drivers are displaying signs of tiredness, and the current research takes a similar approach, monitoring eye and head movement that deviates from normal driving behaviors. The algorithms can detect the duration of distraction using on onboard cameras and will alert the driver to their negligence, for example if they are looking down at their phone for too long or reaching for something in the back seat. The aim is to make the software capable of interacting with connected cars to activate safety features, such as temporary autopilot or collision control systems. Driven by statistics that suggest 75 percent of global traffic accidents are caused by distracted drivers, the team are now aiming to bring all the various detection technologies into one system that can be fitted into vehicles.
There are growing examples of using machine learning to develop image recognition technologies, whether that’s used to improve the efficacy of smart homes or to augment surveillance techniques. In what other contexts could we see this technology used?