Research on Road Scene Understanding of Autonomous Vehicles Based on Multi-Task Learning
Jinghua Guo, Jingyao Wang, Huinian Wang, Baoping Xiao, Zhifei He, Lubin Li
Abstract
Road scene understanding is crucial to the safe driving of autonomous vehicles. Comprehensive road scene understanding requires a visual perception system to deal with a large number of tasks at the same time, which needs a perception model with a small size, fast speed, and high accuracy. As multi-task learning has evident advantages in performance and computational resources, in this paper, a multi-task model YOLO-Object, Drivable Area, and Lane Line Detection (YOLO-ODL) based on hard parameter sharing is proposed to realize joint and efficient detection of traffic objects, drivable areas, and lane lines. In order to balance tasks of YOLO-ODL, a weight balancing strategy is introduced so that the weight parameters of the model can be automatically adjusted during training, and a Mosaic migration optimization scheme is adopted to improve the evaluation indicators of the model. Our YOLO-ODL model performs well on the challenging BDD100K dataset, achieving the state of the art in terms of accuracy and computational efficiency.