Tongji Master's student wins China's first CVPR Best Student Paper Award
On June 22, the IEEE International Conference on Computer Vision and Pattern Recognition (CVPR) 2022 announced the best student paper and other awards, and researchers from Tongji University and Alibaba won the only best student paper award.
The first author of the paper is Chen Hansheng, a Class of 2020 master's student at the Institute of Intelligent Vehicles and Shanghai Science Center for Autonomous Intelligent Unmanned Systems, School of Automotive Sciences, Tongji University, under the supervision of Professor Xiong Lu, and the corresponding authors are Wei Tian, an assistant professor at the School of Automotive Sciences, Tongji University, and Dr. Pi Chao Wang at Alibaba.
This is also the first time that the first author of the winning paper is from a Chinese university since CVPR established the Best Student Paper Award in 2001.
Accurate pose estimation of objects has been a challenge in the field of autonomous driving.
This paper, entitled "EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation," addresses the problem of 6-degree-of-freedom object pose estimation in monocular images. The paper, entitled "EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation", proposes a generalized probabilistic n-point perspective (PnP) algorithm for the 6-degree-of-freedom pose estimation problem in monocular images, which can be inserted as a differentiable module into the depth network to output the pose on SE3 manifold, and achieves end-to-end object pose learning by minimizing the KL dispersion between the predicted pose and the target pose distribution.
EPro-PnP combines the advantages of traditional geometric vision and cutting-edge deep learning fields to improve the accuracy of pose estimation for both the 6-degree-of-freedom object pose estimation task on the LineMOD dataset and the 3D target detection task on the nuScenes dataset, and has a wide range of applications in robotics, autonomous driving, and other fields.
Professor Xiong Lu, the mentor, said that the Tongji students won this important award at the international top conference, which is a reflection of the results of the interdisciplinary cross-collaboration of Tongji University's vehicle engineering, intelligent science and technology, and the crystallization of the university-enterprise industry-university-research cooperation. In the future, the team will continue to promote multidisciplinary cross-science research, deepen the integration of industry and education talent training, and strive to contribute to the realization of high-level scientific and technological self-sufficiency in China.
