Yannick Verdie

Office BC 308, CVLab, EPFL
Station 14, CH-1015, Lausanne

Phone: +41 (0)21 69 37XXX
E-mail: yannick...@epfl.ch
Curriculum vitae

I am Postdoc at the Computer Vision Lab of EPFL, Lausanne, Switzerland, and a experienced research fellow for the CERN, working on the EDUSAFE Marie Curie european project.

My research interests mainly focus on robust keypoint detectors for augmented reality and robotic vision applications in industrial contexts.

About me

I am particularly interested in new challenge, and I fell confortable learning new skills. My predilected domain was computer graphic during my PhD (see my Siggraph 2015 paper), and I start working on computer vision and machine learning problems this year (see my CVPR 2015 paper). All my researches have a stong focus on the applicability of the method to real-case scenario.


I obtained my Ph.D. in Computer Science at INRIA, France under the supervision of Dr. Florent Lafarge and Prof. Josiane Zerubia. My Ph.D. topic was 'Urban scene modeling from airborne data'.

I received two Master degree: a M.Sc. degree on Computer science from Virginia Tech (Virginia, USA) in 2010 and the Diplôme d'Ingénieur Généraliste from the french engineer school Telecom SudParis (France) in 2008.

Research projects

2013 - 2014

Reliable features for augmented reality

Keypoint detection and matching is an essential tool to address many Computer Vision problems such as image retrieval, object tracking, and image registration. We successfully apply our methods indoor in the Atlas detector of CERN (project EDUSAFE), and outdoor with a cellphone implementation (project MAGELLAN).

IEEE CVPR 2015, Verdie Yannick, Moo Yi Kwang, Fua Pascal, Lepetit Vincent (project page)
TILDE: A Temporally Invariant Learned DEtector

Sample Logos


3D modeling of urban scene (part of my PhD thesis )

We introduce a novel approach that reconstructs 3D urban scenes in the form of levels of detail (LODs). Our experiments on complex buildings and large scale urban scenes show that our approach generates meaningful LODs while being robust and scalable.

TOG (2015), Verdie, Yannick and Lafarge, Florent and Alliez, P. (view online)
LOD Generation for Urban Scenes
IEEE ICIP 2011, Verdie, Yannick and Lafarge, Florent and Zerubia, Josiane (view online)
Generating compact meshes under planar constraints: an automatic approach for modeling buildings from aerial LiDAR


2D/3D detection of parametric objects (part of my PhD thesis )

Point processes constitute a natural extension of Markov random fields (MRF), designed to handle parametric objects. The performances of the existing samplers are limited in terms of computation time and convergence stability, especially on large scenes. We propose a new sampling procedure based on a Monte Carlo formalism.

IJCV (2013), Verdie, Yannick and Lafarge, Florent (view online)
Detecting parametric objects in large scenes by Monte Carlo sampling
ECCV 2012, Verdie, Yannick and Lafarge, Florent (view online)
Efficient Monte Carlo Sampler for Detecting Parametric Objects in Large Scenes

Previous publications

ICLS 2010,Workshop 6 Miller, C. Verdie, Y. Quek, F. Ehrich, R. (view online)
Physical-Gestural-Parametric: Learning through graceful transition
SIGGRAPH 2010,Workshop Quek, F., Miller, C., Joshi, A., Verdie, Y., Ehrich, R., Evans, M., Chu Yew Yee, S. and Chakraborty, P.
TanTab, A Tangram Tabletop System Technology to Facilitate Collaborative, Co-Constructive Learning: A Multi-Touch, Tangible User Interface for PreK-2 Mathematics
ICMI-MLMI 2009 Verdie, Y. Fang, B. and Quek, F. (view online)
MirrorTrack - Tracking with Reflection - Comparison with Top-Down Approach

Codes and registered softwares

  • Real-time 3D tracking with Descriptor Fields

  • Code for real-time tracking of planar targets using Desriptor Fields. It includes a library of utilities for computing the Descriptor Fields and a simple real-time tracking demo. (view video) (from this paper)

  • Registred softwares

  • Acute3D Version 1.0 (Febuary 2013): Software for 3D mesh labeling and semantics. (view video) (from this paper)
    Kazoe Version 1.0 (September 2012): Software to count objects from large pictures. (from this paper)
    Midori Version 1.0 (September 2012): Software to detect Trees from large 3D point clouds. (from this paper)
    Nagare Version 1.0 (September 2012): Software to extract lineic structures from larges pictures. (from this paper)

    Misc - old projects


    Surface Gesture and Object Tracking on Tabletop Devices (Master Thesis)

    In my Master thesis, we are interested in the use of tabletop surfaces for interactive manipulations. We focus on the implementation of Image Processing algorithms and techniques in two projects ex- ploiting a horizontal surface: ‘Tangram Project’ and ‘MirrorTrack’. The ‘Tangram Project’ studies children’s mathematical skills when manipulating geometrical shapes. The ‘MirrorTrack Project’ uses a horizontal surface with two side-mounted cameras to track fingertips.


    Bluetooth Electronic bracelet....

    My Enginner 6-month project, about the dev. of a bluetooth bracelet for augmented reality.

    [ Back to Top ]