Peiyun Hu

I am now a Software Engineer at Argo AI working on self-driving cars. I recently completed my Ph.D. in Robotics at Carnegie Mellon University. My thesis can be found here.


My research focuses on computer vision, often motivated by the task of developing perception systems for autonomous robots. My work uses machine learning and deep learning to improve the robustness and scalability of learning-based perception, often led by three questions:

  • What is a good output representation of perception for supporting downstream applications?
  • What is a good input representation of sensor data for solving perception tasks?
  • How can we learn more with less human supervision?
  • Below are some of the projects where we delve into these questions.


    Safe Local Motion Planning with Self-Supervised Freespace Forecasting
    Peiyun Hu, Aaron Huang, John Dolan, David Held, Deva Ramanan
    Computer Vision and Pattern Recognition (CVPR), 2021
    paper / project / poster / talk / code


    Active Perception using Light Curtains for Autonomous Driving
    Siddharth Ancha, Yaadhav Raaj, Peiyun Hu, Srinivasa Narasimhan, David Held
    European Conference on Computer Vision (ECCV), 2020
    (Spotlight Presentation)
    paper / project / slides / talk / code


    What You See is What You Get: Exploiting Visibility for 3D Object Detection
    Peiyun Hu, Jason Ziglar, David Held, Deva Ramanan
    Computer Vision and Pattern Recognition (CVPR), 2020
    (Oral Presentation)
    paper / project / slides / talk / demo / code

    We exploit often overlooked freespace in LiDAR-based 3D object detection.


    Learning to Optimally Segment Point Clouds
    Peiyun Hu, David Held*, Deva Ramanan*
    IEEE Robotics and Automation Letters (RA-L) and ICRA, 2020
    paper / project / slides / talk / demo / code

    We marry graph search with learning for point cloud optimal segmentation.


    Recognizing Tiny Faces
    Siva Chaitanya Mynepalli, Peiyun Hu, Deva Ramanan
    Computer Vision and Pattern Recognition Workshops (CVPR-W), 2019


    Inferring Distributions Over Depth from a Single Image
    Gengshan Yang, Peiyun Hu, Deva Ramanan
    IEEE International Conference on Intelligent Robots and Systems (IROS), 2019
    paper / project / slides / code


    Active Learning with Partial Feedback
    Peiyun Hu, Zack C. Lipton, Anima Anandkumar, Deva Ramanan
    International Conference on Learning Representations (ICLR), 2019
    paper / poster / code

    We formulate AL as a 20Q game between evolving models and human.


    Camera-based Semantic Enhanced Vehicle Segmentation for Planar LIDAR
    Chen Fu, Peiyun Hu, Chiyu Dong, Christoph Mertz, John Dolan
    International Conference on Intelligent Transportation Systems (ISTC), 2018


    Comparing Apples and Oranges: Off-Road Pedestrian Detection on the NREC Agricultural Person-Detection Dataset
    Zachary Pezzementi, Trenton Tabor, Peiyun Hu, Jonathan K. Chang, Deva Ramanan, Carl Wellington, Benzun P. Wisely Babu, Herman Herman
    Journal of Field Robotics (JFR), 2018
    paper / project / video

    Unconstrained Face Detection and Open-Set Face Recognition Challenge
    Manuel Gunther, Peiyun Hu, Christian Herrmann, Chi-Ho Chan, Min Jiang, Shufan Yang, Akshay Raj Dhamija, Deva Ramanan, Jurgen Beyerer, Josef Kittler, Mohamad Al Jazaery, Mohammad Iqbal Nouyed, Guodong Guo, Cezary Stankiewicz, Terrance E Boult
    IEEE International Joint Conference on Biometrics (IJCB), 2017


    Finding Tiny Faces
    Peiyun Hu, Deva Ramanan
    Computer Vision and Pattern Recognition (CVPR), 2017
    paper / project / video / poster / press / code

    We address key challenges in detecting tiny objects with neural nets.


    Bottom-Up and Top-Down Reasoning with Hierarchical Rectified Gaussians
    Peiyun Hu, Deva Ramanan
    Computer Vision and Pattern Recognition (CVPR), 2016
    (Spotlight Presentation)
    paper / project / ext. abstract / slides / talk / poster / code

    We derive architectures to endow neural nets with top-down reasoning.

    webpage design