Jingyu Song

Jingyu Song | 宋靖宇 - PhD Candidate @ U-M Robotics

prof_pic.jpg

Monterey Bay, CA, 2024.05

I am a final-year PhD candidate at the University of Michigan Robotics Department, working with Prof. Katie Skinner at the Field Robotics Group and the Ford Center for Autonomous Vehicle. Before that, I graduated from the University of Michigan with an M.S. in Electrical and Computer Engineering in 2022, where I worked with Prof. Maani Ghaffari. I obtained my B.E. in Electronics and Electrical Engineering jointly from the University of Electronic Science and Technology of China and the University of Glasgow in 2020.

I was honored to receive the Rackham Predoctoral Fellowship in 2025 and be named a Qualcomm Innovation Fellowship (QIF) Finalist in 2024. I was a research intern at NVIDIA in 2024, a research intern at Bosch Corporate Research in 2020, and an undergraduate research intern at Intel in 2018.

I am interested in the intersection of robotics, computer vision, and deep learning, with a focus on multi-modal perception, state estimation and mapping for 3D scene understanding. I am also interested in the application of these technologies in autonomous driving and field robotics.

news

February 2026 The workshop I co-organized, ICRA 2026: S2S Perception, has been accepted! I will also present my work DriveCritic there.
September 2025 I will attend OCEANS 2025 and will chair two technical sessions! Looking forward to seeing you there!
May 2025 I joined the Autonomous Vehicles Applied Research Group led by Dr. Jose M. Alvarez in NVIDIA as a research intern in summer 2025! 🚗
April 2025 Our latest work, OceanSim: A GPU-Accelerated Underwater Robot Perception Simulation Framework, has been released. Excited to see it’s featured by NVIDIA, and many others! Check it out on LinkedIn and X/Twitter!
April 2025 I will attend ICRA 2025 Doctoral Consortium in Atlanta. I will also present OceanSim there in the late-breaking poster session and the AQ²UASIM workshop!
March 2025 I was selected to receive the Rackham Predoctoral Fellowship! 🎉

selected publications

  1. arXiv
    FishDetector-R1: Unified MLLM-Based Framework with Reinforcement Fine-Tuning for Weakly Supervised Fish Detection, Segmentation, and Counting
    Yi Liu*, Jingyu Song*, Vedanth Kallakuri, and Katherine A Skinner
    arXiv preprint arXiv:2512.05996, 2025
    arXiv Website
  2. ICRA
    DriveCritic: Towards Context-Aware, Human-Aligned Evaluation for Autonomous Driving with Vision-Language Models
    Jingyu Song, Zhenxin Li, Shiyi Lan, Xinglong Sun, Nadine Chang, Maying Shen, Joshua Chen, Katherine A Skinner, and Jose M Alvarez
    IEEE International Conference on Robotics and Automation (ICRA), 2026
    arXiv
  3. IROS
    OceanSim: A GPU-Accelerated Underwater Robot Perception Simulation Framework
    Jingyu Song*, Haoyu Ma*, Onur Bagoren, Advaith V Sethuraman, Yiting Zhang, and Katherine A Skinner
    IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2025
    arXiv Website
  4. WACV
    MemFusionMap: Working Memory Fusion for Online Vectorized HD Map Construction
    Jingyu Song, Xudong Chen, Liupei Lu, Jie Li, and Katherine A Skinner
    IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2025
    arXiv Website
  5. IROS
    TURTLMap: Real-time Localization and Dense Mapping of Low-texture Underwater Environments with a Low-cost Unmanned Underwater Vehicle
    Jingyu Song*, Onur Bagoren*, Razan Andigani, Advaith Venkatramanan Sethuraman, and Katherine A. Skinner
    IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024
    arXiv Website
  6. CVPR
    CRKD: Enhanced Camera-Radar Object Detection with Cross-modality Knowledge Distillation
    Lingjun Zhao*, Jingyu Song*, and Katherine A Skinner
    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024
    arXiv Website
  7. ICRA
    LiRaFusion: Deep Adaptive LiDAR-Radar Fusion for 3D Object Detection
    Jingyu Song, Lingjun Zhao, and Katherine A Skinner
    IEEE International Conference on Robotics and Automation (ICRA), 2024
    arXiv Website