Xiaohan "Bruce" Nie    聂潇寒

Phd Candidate
Advisor: Prof.Song-chun Zhu
Center for Vision, Cognition, Learning and Autonomy
Statistical Department
office: 9107 Bolter Hall, UCLA


  • Education
    • M.S. in Computer Science, Beijing Institute of Technology, Beijing, China, March 2012
    • B.S. in Computer Science, Zhengzhou University, Zhengzhou, China, July 2009
  • Research Experiences
    • Visiting Student, Lotus Hill Institute, Wuhan, China, June 2010 - July 2012
    • Research Assistant, Beijing Institute of Technology, Beijing, China, September 2009 - March 2010

Here is my CV.

Research Interests

  • Computer Vision, Machine Learning.
  • Human Action Recognition, Attribute Recognition, Pose Estimation

Research Projects

  • Joint Parsing of Human Attributes and Pose In this work we infer the human body pose and human attributes simultaneously using a hierarchical representation. In inference, We first generate human part proposals with attributes using a DCNN model, and then the proposals are integrated with an attributed hierarchical structure which models attributes and parts together at different scales.
  • Joint Action Recognition and Pose Estimation from Video. We represent human action and poses in a same framework by using hierarchical model. The action is decomposed into mid-resolution parts and then high-resolution parts. The movement of human body in the video is captured by temporal transition of mid-level parts.
  • Multi-view Action Modeling and Detection. We aim to detect and recognize actions from multiple views. A spatial-temporal structure is used to represent actions from multiple viewpoints. We leverage 3D skeleton data for training but in inference we don't need skeleton data. See Demo
  • Animated Pose Templates for Modeling and Detecting Human Actions.Each action is repesented by several key-poses each of which has appearance template and motion template. We use HMM to model the temporal transitions between poses. The inference is conducted efficiently by Dynamic Programming
  • Arbitrary view license plate recognition. A three layer part-based model is used to represent the license plate. The first layer represents rough plate, second level represents the characters inside the plate and third level represents the cells which construct the character. Every part deforms according to the affine transformation under different viewpoint. Latent-SVM is used for trainning and Dynamic programming is used for inference. See Demo
  • Publications

  • Attribute And-Or Grammar for Joint Parsing of Human Attributes, Part and Pose
  • Seyoung Park, Bruce Xiaohan Nie, Song-Chun Zhu.
    arXiv:1605.02112, pdf
  • Joint Action Recognition and Pose Estimation From Video
  • Bruce Xiaohan Nie, Caiming Xiong, Song-Chun Zhu.
    Computer Vision and Pattern Recognition, 2015, Boston, massachusetts, pdf
  • Cross-view Action Modeling, Learning and Recognition
  • Jiang Wang, Bruce Xiaohan Nie, Yin Xia, Ying Wu, Song-Chun Zhu.
    Computer Vision and Pattern Recognition, 2014, Columbus, Ohio, pdf
  • Animated Pose Templates for Modelling and Detecting Human Actions.
  •    Benjamin Yao, Bruce Xiaohan Nie, Zichen Liu, and Song-Chun Zhu.
        IEEE Trans on Pattern Analysis and Machine Intelligence, 2013, pdf
  • A benchmark for interactive image segmentation algorithms.
  •    Yibiao Zhao, Xiaohan Nie, Yanbiao Duan, Yaping Huang, SiweiLuo
       IEEE Workshop on Person-Oriented Vision (POV), 33-38, 2011, pdf
  • Experts-Shift: Learning active spatial classification experts for keyframe-based video segmentation.
  •   Yibiao Zhao, Yanbiao Duan, Xiaohan Nie, Yaping Huang, Siwei Luo
       IEEE Workshop on Applications of Computer Vision (WACV), 622-627, 2011, pdf

    Software and Data

    Please see the project page.


    Teaching Assitant, Stat102A Introduction to Computational Statistics with R, Fall, 2016
    Special Reader, Stat232B Statistical Computing and Inference in Vision and Image Science, Spring, 2016
    Teaching Assitant, Stat102B Computation and Optimization for Statistics, Winter, 2016
    Special Reader, Stat202C Monte Carlo Methods for Optimization, Spring, 2015