About Me

I am currently a third-year Master's student in Computer Science at ShanghaiTech University. I am interested in Computer Vision (depth estimation and semantic segmentation), Robotics and SLAM (visual odometry and 3D reconstruction). Before joining ShanghaiTech University, I got my bachelor degree from Hefei University of Technology with highest honor. My resume can be downloaded [here].

I’m looking for PhD positions starting Spring 2021. Please contact me if you are interested in hiring me. Many thanks!

Education

  • September 2017 - Present
    ShanghaiTech University
    Computer Science - Master of Science (MS)
    Shanghai - China
  • September 2013 - June 2017
    Hefei University of Technology
    Computer Science and Technology - Bachelor of Engineering (BE)
    XuanCheng, Anhui - China
    • Team leader of 2D Simulation League Team of HfutEngine Group
    • Undergraduate Thesis with Prof.Baofu Fang in 2D SLAM and Navigation in Indoor Environments.
    • First Prize Scholarship of HeFei University of Technology

Work Experience

  • November 2019 - Present
    Research Intern
    Amazon Shanghai AI Lab Shanghai - China
    • Establishing semantic segmentation and depth estimation model to Amazon GluonCV library
    • Trying to speed up the model via replacing existing model to a lightweight network for the low cost devices

  • June 2016 - September 2016
    Visiting Student
    Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen - China

Projects

  • Robot Pose Estimation for Omni-directional Cameras
    Haofei Kuang, Qingwen Xu, Xiaoling Long and Sören Schwertfeger

    • Extracted motion of pixels between two frames via improved Fourier-Mellin invariant (iFMI) algorithm.
    • Estimated the full 6 DoF 3D transform, up to an unknown scale factor via fitting the motion of the pixels in the panoramic images to two sinusoidal functions.
    [IROS'19]
  • Depth Estimation by using Deep Neural Network
    Haofei Kuang, Qingwen Xu and Sören Schwertfeger

    • Established a Deep Fully Convolutional Residual Network for predicting a full resolution depth map from a single monocular image.
    • Developed a spherical convolution to deal with distortion problems of panoramic images.
    [Workshop on ICRA'19]
  • Image Registration via Fourier-Mellin transform (FMT) Algorithm
    Haofei Kuang and Qingwen Xu

    • Course Project of CS270 Digital Image Processing
    • Solved 2D image registration problem between two related images via FMT algorithm.
    • Implemented FMT algorithm with omnidirectional camera model to found the correspondences between two continue frames.
    [report] [code]
  • Image Registration by using Gauss-Newton Method
    Haofei Kuang

    • Course Project of SI211 Numerical Analysis
    • Formulated 2D image registation problem as a least-squared problem and written the objective function.
    • Solved a image registration problem by using Gauss-Newton method.
    [report] [code]
  • Full-Speed Indoor Navigation with Jackal Robot
    Haofei Kuang, Wang Lei, Li Ziyue and Zeng Xiangchen

    • Course Project of CS283 Robotics
    • Mounted a Velodyne VLP-16 laser scanner on Jackal robot for the environmental perception.
    • Implemented the PID controller with Jackal robot to solve a indoor local navigation task.
    [report] [code] [webpage]

Publications

  • IROS'19
    Pose Estimation for Omni-directional Cameras using Sinusoid Fitting
    Haofei Kuang, Qingwen Xu, Xiaoling Long and Sören Schwertfeger

    [pdf]
  • ROBIO'19
    Fast 2D Map Matching Based on Area Graphs
    Jiawei Hou, Haofei Kuang and Sören Schwertfeger

    [pdf]
  • Workshop on Underwater Robotics Perception, ICRA'19
    Depth Estimation on Underwater Omni-directional Images Using a Deep Neural Network
    Haofei Kuang, Qingwen Xu and Sören Schwertfeger

    [pdf] [poster] [best paper award]
  • ICARCV'18
    Fast Gaussian Process Occupancy Maps
    Yuan Yijun, Haofei Kuang and Sören Schwertfeger

    [pdf] [code]