WORKSHOP DETAILS

WS10: Scene Understanding for Automated Driving Systems
( CFP )

Workshop Code: e434d

Time: 8:30-12:00

Chair

• Jianru Xue
  Affiliation: Xi’an Jiaotong University
  E-mail: jrxue@mail.xjtu.edu.cn

Cochair

• Ming Yang
  Affiliation: Shanghai Jiaotong University
  E-mail: mingyang@sjtu.edu.cn

• Huijing Zhao
  Affiliation: Peking University
  E-mail: zhaohj@cis.pku.edu.cn

• Jianwu Fang
  Affiliation: Xi’an Jiaotong University
  E-mail: fangjianwu@mail.xjtu.edu.cn

Scope and Goals

Automated driving systems has become one of the most exciting and important innovations in transportation history. Scene understanding, at the core of an automated driving systems, performs sensing, comprehending, predicting surrounding traffic scene of the ego-vehicle. Specifically, scene understanding should know the geometry/topology of traffic scene, participants’ behaviour (pedestrian, vehicle, cyclist, road, etc.), as well as their spatio-temporal evolution, implicitly contained in the sensing data. With the scene understanding, the autonomous vehicle can facilitate the semantical reasoning of traffic scene, intention prediction of participants, accurate motion planning, as well as other promising applications.

Most recently, benefiting from rapid improvement of sensing technologies, scene understanding of autonomous driving could integrate information of different sensors, including vision, lidar, radar, GPS/IMU, and so forth, together to achieve a better and more accurate understanding. Additionally, in cognitive science, understanding the surrounding scene involves human experience and memory, visual attention, brain reasoning, and so on. How to facilitate autonomous vehicle can comprehend the traffic scene like human beings? This also is a challenging and frontier problem in this field. To help make further progress in this field, we propose to invite experts in this domain to participate in discussions, and showcase their latest innovations/ideas.

Topics of Interest

• Pedestrian/vehicle detection, tracking, recognition
• Motion intention prediction of traffic participants
• Motion representation, inference and classification of traffic participants
• Road scene geometry/topology representation and reasoning
• Traffic event representation, detection, classification
• Multi-sensor fusion for autonomous driving
• Brain-inspired traffic situation understanding
• Deep learning for understanding the traffic scene

Schedule

• Visual SLAM and Its Application
  Guofeng Zhang
  Zhejiang University
  zhangguofeng@cad.zju.edu.cn（40min）

• Visual Saliency Detection and Sequence Learning for Autonomous Driving
  Xin Xu
  National University of Defense Technology
  xinxu@nudt.edu.cn（40min）