Ph.D., Robotics Institute
School of Computer Science
Carnegie Mellon University

Office: Smith Hall 210

 total page visits since 06/20/2017
Free Web Counters unique visits since 06/20/2017

Xinshuo Weng is a Ph.D. student (2018-) at Robotics Institute of Carnegie Mellon University (CMU) supervised by Kris Kitani. She received Masters (2016-17) also at CMU, where she worked with Yaser Sheikh and Kris Kitani. Prior to CMU, she worked at Oculus Research Pittsburgh (now Facebook Reality Lab) as a research engineer. Her Bachelor's degree was received from Wuhan University. Her primary research interest lies in 3D computer vision and Graph Neural Networks for autonomous systems. She was awarded a Qualcomm Innovation Fellowship for 2020-2021.

Fields: Computer Vision, Machine Learning, Robotics, Multimedia
Topics: 3D Computer Vision, Autonomous Driving, Graph Neural Networks, Generative Modeling, Video Analysis

Google Scholar  /  DBLP  /  Semantic Scholar  /  GitHub  /  LinkedIn  /  Twitter  /  ResearchGate
  • 02/2021 - One paper accepted at RA-L + ICRA 2021
  • 12/2020 - Invited speaker at Wayve and Computer Vision Talks
  • 12/2020 - Honored to receive the Outstanding Reviewer Award at ACCV 2020
  • 10/2020 - Papers accepted at CoRL 2020, WACV 2021, ISARC 2020
  • 08/2020 - Honored to receive the Qualcomm Innovation Fellowship 2020
  • 08/2020 - Keynote speaker at ECCV 2020 Workshop on Benchmarking Trajectory Forecasting Models [Slides]
  • 08/2020 - Organizing NeurIPS 2020 Workshop on Machine Learning for Autonomous Driving
  • 08/2020 - Four (one oral, three spotlight) papers accepted at ECCV 2020 Workshops
  • 06/2020 - Two papers accepted at IROS 2020
  • 06/2020 - Keynote Speaker at CVPR 2020 Workshop on Scalibility in Autonomous Driving [Slides] [Video]
  • 04/2020 - One paper accepted at TPAMI 2020
  • 03/2020 - One paper accepted at CVPR 2020
  • See More
  • 08/2019 - One paper accepted at ICCV Workshops 2019
  • 06/2019 - We release the code for our 3D MOT paper here
  • 06/2019 - Three papers accepted at BMVC 2019, ACMMM 2019, IROS 2019
  • 01/2018 - One paper accepted at CVPR 2018
  • 10/2017 - One paper accepted at WACV 2018
  • Parallelized 3D Tracking and Forecasting with Graph Neural Networks and Diversity Sampling
    Xinshuo Weng*, Ye Yuan*, Kris Kitani
    Robotics and Automation Letters (RA-L), 2021
    with presentation at IEEE International Conference on Robotics and Automation (ICRA), 2021
    PDF | Code | Website | BibTex
    The first parallelized 3D MOT and trajectory forecasting method with object interaction modeling and diverse trajectory samples

    Inverting the Pose Forecasting Pipeline with SPF2: Sequential Pointcloud Forecasting for Sequential Pose Forecasting
    Xinshuo Weng, Jianren Wang, Sergey Levine, Kris Kitani, Nick Rhinehart
    Conference on Robot Learning (CoRL), 2020
    PDF | Code | Demo | Website | Slides | BibTex | Supp
    By learning to forecast future LiDAR point clouds, we build a forecast-then-detect pipeline for pose forecasting by reversing the steps of detection and forecasting

    GNN3DMOT: Graph Neural Network for 3D Multi-Object Tracking with 2D-3D Multi-Feature Learning
    Xinshuo Weng, Yongxin Wang, Yunze Man, Kris Kitani
    IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2020
    PDF | Code | Demo | Website | Slides | BibTex
    The first multi-object tracking method that leverages Graph Neural Network for object interaction modeling

    3D Multi-Object Tracking: A Baseline and New Evaluation Metrics
    Xinshuo Weng, Jianren Wang, David Held, Kris Kitani
    IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020
    PDF | Code | Demo | Website | Slides | BibTex
    A 3D multi-object tracker that achieves state-of-the-art performance with the fastest speed

    Monocular 3D Object Detection with Pseudo-LiDAR Point Cloud
    Xinshuo Weng, Kris Kitani
    IEEE International Conference on Computer Vision (ICCV) Workshops, 2019.
    PDF | Code | Poster | BibTex
    By projecting the 2D image to a pseudo-LiDAR point cloud representation, our monocular 3D detection pipeline quadruples the performance over prior art

    Back to top