Xinshuo Weng

Ph.D.
The Robotics Institute
School of Computer Science
Carnegie Mellon University
Pittsburgh, PA 15213, USA

Office: 1502F NSH
Email: xinshuow@cs.cmu.edu

 page loads since 06/20/2017

Github LinkedIn CMU


Brief Bio

I am a first-year Ph.D. student (2018-) at the Robotics Institute of Carnegie Mellon University supervised by Kris Kitani. I received my Masters (2016-17) at the Robotics Institute as well, where I am working with Yaser Sheikh and Kris Kitani. Before starting my Ph.D. program at CMU, I worked at Oculus Research Pittsburgh (Facebook Reality Lab) as a research engineer. I spent a wonderful summer (2016) working with Alan Yuille at the Johns Hopkins University as a summer intern. When I was an undergradute, I've studied in School of Computer Science, University College Dublin as an exchange student in Ireland. My Bachelor's degree is received from the School of Electronic Information at Wuhan University in China.

See here for my resume.


Research Interests

Computer vision and machine learning: understanding humans from images and video, machine learning for vision, image motion and tracking, 3D vision, and the intersection of reinforcement learning and vision.


News

Aug. 2018 -- Joined CMU Robotics Institute as a Ph.D. student.
Feb. 2018 -- Joined Oculus Research Pittsburgh as a research engineer.
Jan. 2018 -- One paper accepted to CVPR 2018!
Oct. 2017 -- One paper accepted to WACV 2018!
May. 2017 -- Joined Facebook as a research intern.
Aug. 2016 -- Started the master program in computer vision (MSCV) In Robotics Institute at CMU.
Jun. 2016 -- Joined Alan Yuille's group as a summer intern.


Research Projects


GroundNet: Segmentation-Aware Monocular Ground Plane Estimation with Geometric Consistency
[PDF]

Yunze Man, Xinshuo Weng, Kris Kitani

arXiv preprint 2018

Supervision-by-Registration: An Unsupervised Approach to Improve the Precision of Facial Landmark Detectors
[PDF]

Xuanyi Dong, Shoou-I Yu, Xinshuo Weng, Shi-en Wei, Yi Yang, Yaser Sheikh

IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018

Rotational Rectification Network: Enabling Pedestrian Detection for Mobile Vision
[PDF][Poster][Oral]

Xinshuo Weng, Shangxuan Wu, Fares Beainy, Kris Kitani

IEEE Winter Conference on Applications of Computer Vision (WACV), 2018

Visual Compiler: Synthesizing a Scene-Specific Pedestrian Detector and Pose Estimator
[arXiv]

Namhoon Lee, Xinshuo Weng, Vishnu Naresh Boddeti, Yu Zhang, Fares Beainy, Kris Kitani, Takeo Kanade

ArXiv preprint arXiv:1612.05234

Learning Coherency for Human Segmentation with Skeleton
[Demo][Poster][Web Page]

Xinshuo Weng, Shangxuan Wu, Donghyun Yoo, Kris Kitani

Related Papers[ICCV2015][CVPR2016][CVPR2015][CVPR2016]
Unsupervised Learning of Object Semantic Parts from Internal States of CNNs by Population Encoding

Xinshuo Weng*, Yingda Xia*, Fanchao Qi*, Alan Yuille

Related Papers[arXiv2015]


Bayesian Sparse Regression for Photometric Stereo Reconstruction
[Presentation][Code]

Xinshuo Weng, Lei Yu

Related Papers[TPAMI2014]


Publications

1. Y. Man, X. Weng, K. Kitani, "GroundNet: Segmentation-Aware Monocular Ground Plane Estimation with Geometric Consistency", ArXiv preprint arXiv:1811.07222.

1. X. Dong, S. Yu, X. Weng, S. Wei, Y. Yang, Y. Sheikh, "Supervision-by-Registration: An Unsupervised Approach to Improve the Precision of Facial Landmark Detectors", Computer Vision and Pattern Recognition (CVPR), 2018.

2. X. Weng, S. Wu, F. Beainy, K. Kitani, "Rotational Rectification Network: Enabling Pedestrian Detection for Mobile Vision", IEEE Winter Conference on Applications of Computer Vision (WACV), 2018

3. N. Lee, X. Weng, V. Boddeti, Y. Zhang, F. Beainy, K. Kitani, T. Kanade, "Visual Compiler: Synthesizing a Pedestrian Pose Estimator from a Single Image", ArXiv preprint arXiv:1612.05234.


Teaching

Geometry-Based Methods in Computer Vision (16-822), CMU
Teaching Assistant (TA) with Martial Hebert
Fall 2018


Code

1. Xinshuo's Toolbox: A self-contained structured toolbox for computer vision and machine learning (deep learning). It's written in python, matlab, c++ and lua, containing extensive libraries for I/O stream, image & video processing, visualization for convenience.

2. CNN Monitor: A very lightweight deep learning tool for monitoring data flow, parameter size and their corresponding memory usage throughout CNN. This tool doesn't need any powerful computational resource (eg. GPU). And it's very easy to use since it follows many similar rules in popular deep learning frameworks (Caffe, Tensorflow, Torch)


Awards and Honors

  • Outstanding Graduate Award, Wuhan University, 2016.
  • Wuhan University Scholarship (4%), 2013, 2015, 2016.
  • CSC (China Scholarship Council) Scholarship (1%), 2015.
  • Yang Gui Scholarship (4%), Wuhan University, 2015.
  • Undergraduate Research Fellowship, Wuhan University, 2014, 2015.
  • China National Scholarship (1%), 2014.


Professional Service

  • Conference Reviewer: CVPR, ACCV.
  • Journal Reviewer: TCSVT.


Free Web Counters unique visitors since 06/20/2017