DriveSense

Abstract

Modern self-driving systems seek to remove control from the human driver, placing them in a monitoring role that humans do not perform well in. However, this same technology can be utilized to improve existing driver’s skills using a much cheaper piece of hardware already present in many vehicles, the dashcam. Our system uses a vision model and a set of heuristics to analyze this footage and provide driving statistics to the user, helping them to gain a holistic view of their driving patterns and trends. This information helps them to reflect and create actionable goals to improve their driving in the future. For instance, a driver that is consistently being passed by others when they are in the leftmost lane should consider moving over to enable the natural flow of traffic, a trend our system can identify.