Computer Vision News - November 2018

12 Computer Vision News Distracted Driver Detection with Deep Learning Every month, Computer Vision News reviews a successful project. Our main purpose is to show how diverse image processing techniques contribute to solving technical challenges and physical difficulties. This month we review RSIP Vision ’s solution to track driver attention: Distracted Driver Detection with Deep Learning . RSIP Vision’s engineers can assist you in countless application fields. Project More than one million deaths occur in the world every year as a consequence of car accidents. The number of injuries and severe harm is on the order of tens of million per year. It has been found that about one fifth of these accidents are due to distracted driving , defined as driving while performing other activities that take the driver’s attention away from driving. Traditional reasons for distraction are somnolence, looking at scenery, heated discussions with passengers of the car and operating various car/radio/CD commands. The massive availability of smartphones added new and deadly threats, under the form of talking on the phone, texting and reading social media - while driving. Talking on a cell phone is said to quadruple the risk of crashing . Among the many projects conducted by RSIP Vision in the automotive field, today we will discuss the development of a system able to accurately detect when a driver is entering a state of distraction . A conveniently placed dashboard camera can detect several indicators of the driver’s status and behavior, in particular gaze and movements . Patterns which differ from normal driving are identified in the images coming from the camera: typical patterns are labelled accordingly in view of training the system. The main challenge of doing this is the huge variability of situations : a large set of movements can be made by the driver which are not dangerous; while only those signs which can be considered a clear indicator of fatigue or other distraction should be flagged. This means that a very large amount of data needs to be collected for the training. … RSIP Vision’s deep learning algorithms … provide a great solution to a real world problem! Take us along for your next Deep Learning project!