About Me

I am a first-year Ph.D. in FSM Lab at City University of Hong Kong, advised by Prof. Jianping Wang. I graduated from Beihang University with a master’s degree(Distinction) in Computer Technology in June 2022. I then joined Momenta as a Perception Algorithm Engineer. I have a passion for Autonomous Vehicles(AV), focusing on robust and efficient scene understanding algorithms to make AV a reality！

Interests
  • Multi-Object Tracking
  • Sensor Fusion
  • 2D/3D Scene Understanding
  • Autonomous Vehicles and Robots
Education & Work Experience

  • PhD in Computer Science, 2024-Until Now

    City University of Hong Kong (CityU)

  • Senior Perception Algorithm Engineer, 2022-2023

    Momenta, Lidar Perception

  • Msc in Computer Technology, 2019-2022

    Beihang University (BUAA)

News & Updates

  • 2024-05: Our team (FSM Speed) achieved the championship at the 17th F1Tenth Grand Prix! [ project link]

    • Featured Publications

    Zhenchao Ouyang, Xiaoyun Dong, Jiahe Cui, Jianwei Niu, Mohsen Guizani,
    In IEEE Transactions on Intelligent Transportation Systems(T-ITS), 2022
    PV-EncoNet: Fast Object Detection Based on Colored Point Cloud
    PV-EncoNet: Fast Object Detection Based on Colored Point Cloud

    An efficient 3D detection method based on sensor fusion which can reduce the amount of data by 50%.

    PDF Cite Code
    Xiaoyun Dong, Zhenchao Ouyang, Jianwei Niu,
    Outstanding Graduation Thesis in Beihang University (Top 3%)
    Multi-Object Tracking Technology for Autonomous Driving
    Multi-Object Tracking(MOT) Technology for Autonomous Driving

    A real-time and robust MOT system based on different sensors data under different weather scenarios.

    Recent Publications

    PV-EncoNet: Fast Object Detection Based on Colored Point Cloud
    Zhenchao Ouyang, Xiaoyun Dong, Jiahe Cui, Jianwei Niu, Mohsen Guizani, In T-ITS
    PDF Cite Code
    Polarmask-Tracker: Lightweight Multi-Object Tracking and Segmentation Model for Edge Device
    Xiaoyun Dong, Zhenchao Ouyang, Zeling Guo, Jianwei Niu, In ISPA
    PDF Cite
    Fast 3D Point Cloud Target Tracking based on Polar-Voxel Encoding
    Xiaoyun Dong, Zhenchao Ouyang, Changjie Zhang, Jiahe Cui, Jianwei Niu, In SMC
    PDF Cite
    Fast Segmentation-Based Object Tracking Model for Autonomous Vehicles
    Xiaoyun Dong, Jianwei Niu, Jiahe Cui, Zhenchao Ouyang, In ICA3PP
    PDF Cite

    Projects

    *
    All Datasets 3D Detection MOT Autonomous Vehicles
    The 17th F1TENTH Grand Prix (HongKong)
    In both the lap time race and head-to-head competition, our racecar demonstrated superior speed and obstacle avoidance capabilities, leading our team to achieve the championship.
    The 17th F1TENTH Grand Prix (HongKong) Project Code Document
    Geometric Obstacle Detection based on Pointclouds(In Momenta)
  • Created a large-scale dataset of sphere obstacles. Total frames: 33k
  • Proposed a network model for detecting stone pillars and other spheres.
  • The AP for other spheres increased from 60.8% to 71.2%. It has been deployed to real-world cars.
    • Geometric Obstacle Detection based on Pointclouds
    Performance Optimization for Dynamic Object Detection for Autonomous Driving(In Momenta)
  • Built a closed-loop solution of “problem collection-data accumulation-model optimization” from scratch.
  • Contributed more than 12k lines of code, which processed more than 1000w data in total.
  • The data processing speed is increased by 40 times (3k to 120k), and the optimization time is reduced by 75% (2 months to 2 weeks).
    • Performance Optimization for Dynamic Object Detection for Autonomous Driving
    Collected and created a Multi-Object Tracking Dataset in open world scenario
  • Total frames: 7680
  • design a fast annotation algorithm
  • involve 3D pointclouds data, 2D images data, 3D&2D detection ground truth and tracking id
    • Collected and created a Multi-Object Tracking Dataset in open world scenario

    Gallery

    To be done.