Machine Intelligence Laboratory

Cambridge University Department of Engineering

Je Hyeong (John) Hong Je Hyeong Hong

Background - Research - Publications - Teaching

Position: PhD Candidate

E-mail: jhh37 [at]

Office Location: BN4-80

I was a PhD student in the Computer Vision group under Prof. Roberto Cipolla's guidance. My research involved developing algorithms and strategies for low-rank optimization problems in computer vision, namely bundle adjustment in 3D vision. During my PhD, I closely worked with Dr Andrew Fitzgibbon and Dr Christopher Zach. My study was generously funded by Microsoft Research PhD Scholarship and Toshiba Research Europe Studentship.


  1. [10 May 2018] A patch for Ceres Solver implementing VarPro is available here.
  2. [19 May 2018] I will attend a graduation ceremony in Cambridge.
  3. [17 June 2018] I will leave Snap and travel to Salt Lake City for a poster presentation at CVPR.
  4. [25 June 2018] I will be starting my new job at the Korea Institute of Science and Technology (KIST) in Seoul.


I obtained BA and MEng honours degrees from the University of Cambridge. You can find my CV here.

Research Interests

  • Low-rank matrix factorization problems with missing data
  • Structure-from-motion including bundle adjustment
  • Bivariate and other nonlinear optimization problems


  • pOSE: Pseudo Object Space Error for Initialization-free Bundle Adjustment
    J.H. Hong and C. Zach
    2018 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), to appear.
    [ Paper ] [ Supplementary ] [ Repo ] [ Poster ] [ Video ]
  • Revisiting the Variable Projection Method for Separable Nonlinear Least Squares Problems
    J.H. Hong, C. Zach and A.W. Fitzgibbon
    2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
    [ Paper ] [ Supplementary ] [ Repo ] [ Poster ] [ Video ] [ Code ]
  • Projective Bundle Adjustment from Arbitrary Initialization using the Variable Projection Method
    J.H. Hong, C. Zach, A.W. Fitzgibbon and R. Cipolla
    2016 European Conference on Computer Vision (ECCV)
    [ Paper ] [ Supplementary ] [ Repo ] [ Poster ]
  • Secrets of Matrix Factorization: Approximations, Numerics, Manifold Optimization and Random Restarts
    J.H. Hong and A.W. Fitzgibbon
    2015 IEEE International Conference on Computer Vision (ICCV)
    [ Paper ] [ Supplementary ] [ Repo ] [ Poster ] [ Video ]


  • Snap, Los Angeles, USA (2018): I am working with Oliver Woodford and Ed Rosten on a large-scale optimization project.
  • Oculus (Facebook), Redmond, USA (2017): I worked with Alexander Fix on an eye-tracking project, developing an optimization strategy using nonlinear least squares and the implicit function theorem.
  • Toshiba, Cambridge, UK (2015, 2017): I worked with Christopher Zach on a projective bundle adjustment project and a global structure-from-motion strategy using the Variable Projection (VarPro) method.
  • Microsoft, Cambridge, UK (2014): I worked with Andrew Fitzgibbon on a project involving low-rank matrix factorization with missing data.


  • Secrets of Matrix Factorization and its Application to Projective Bundle Adjustment
    KIST IMRC lab (Korea), Dec 2016
    ETRI (Korea), Aug 2016
  • Comparing Matrix Factorization Algorithms on a Level Playing Field [ Video ]
    Microsoft Research Cambridge (CamAIML Workshop), Mar 2016


  • Supervisor (2015-2017) for 3rd-year Cambridge engineering undergraduate students teaching in groups of pairs.
  • Teaching Assistant for a 4th year module 4F13 Machine Learning (2015).
  • Research Assistant for (2013), an online tool to help students prepare for university engineering entrance interviews.