Real-time Implementation of 11-key Pose Estimation for Driver Behavior Analysis

Driver behavior analysis is one of the most important key function in future autonomous driving systems. An important performance indicator here is real-time quality because it is directly related to safety. Therefore, in this paper, definition of 11 key points for real-time driver posture estimation, design of convolutional neural network(CNN) architecture, and its hardware implementation are presented. As a implementation result, we can get under 10ms latency and with 1.25ms input image interval.