Post-doctoral researcher, Computer Vision group, Technical University Munich

Office 02.09.053, Department of Computer Science, Informatik 9, Boltzmannstrasse 3, 85748 Garching, Germany

tel: +49-89-289-17784

@: yvain.queau@tum.de

Short Bio

I'm a post-doctoral researcher at the Computer Vision group of Technical University München, working under the supervision of Prof. Daniel Cremers. Prior to that, I was a research and teaching assistant in INP-ENSEEIHT, University of Toulouse, where I obtained the Ph.D degree in 2015. My CV is available on the right.

My Ph.D dissertation is entitled "Reconstruction tridimensionnelle par stéréophotométrie" (3D-reconstruction by photometric stereo), and can be found here. I received the "Leopold Escande prize" from INP for this dissertation. A short English overview is provided here.

My primary research interest is variational methods for inverse problems in signal and image processing. During the past few years, I've been more specifically tackling computer vision problems such as 3D-reconstruction and reflectance estimation, using photometric techniques such as shape-from-shading and photometric stereo. Now I am willing to use the expertise I gained on variational methods for solving other signal, image and vision problems.

CV_YVAIN_QUEAU.pdf

Publications

News:

  • One paper accepted at ICCVW 2017
  • One paper accepted at CVPR 2017.
  • Three papers accepted at SSVM 2017.

An up-to-date publication list is available here.

publication_list_Yvain_Queau.pdf

Work In Progress

  • Shading-based refinement of RGBD data / MVS results. A recent work showing that a non-convex ADMM-based variational method for shape-from-shading is able to refine depth from an RGBD sensor without any kind of regularization, and to solve dense multi-view stereo using only sparse correspondences. Another recent work where we achieve depth map sup-resolution for RGBD sensors, by resorting to uncalibrated photometric stereo (to be presented in ICCV Workshops).
  • Integration of 2D gradient data into depth map: a survey of existing approaches, new insights about fast least-squares integration or non-convex variational methods for discontinuity-preserving integration.
  • Photometric stereo with nearby point light sources (e.g., LEDs): modeling, calibration and resolution. Check updated code with lots of datasets on GitHub: https://github.com/yqueau/near_ps .

Teaching

I have tought around 300 hours in INP-ENSEEIHT, during my Ph.D ("moniteur" position) and then as a full-time teaching assistant ("ATER" position):

  • Audio, image and video processing (master). Lectures, tutorials and practicals. Creation of new lectures and practicals on variational methods for image segmentation ("snakes"), denoising / inpainting (total variation), and processing ("Poisson image editing").
  • Statistics and probability (bachelor). Lectures, tutorials and practicals. Creation (with J.-D Durou) of new practicals introducing applications to signal and image processing.
  • Computer vision (master). Lectures and practicals. Creation (with J.-D. Durou) of new lectures and practicals on photometric 3D-reconstruction.
  • Android programming (master). Practicals. Creation of new practicals on multimedia streaming.
  • Operational research (bachelor). Lectures, tutorials and practicals.
  • Distributed systems and applications (bachelor). Practicals.

Trainees supervision

Current:

  • Jean Mélou (Ph.D candidate). 2016 - ?. Creation of 3D models for post-production. 50% supervision, with J.-D. Durou.
  • Songyou Peng (MSc candidate). 2017. Real-time depth refinement with color photometric stereo.
  • Qassim Akhtar (MSc candidate). 2017. A portable high resolution 3D-reconstruction system.
  • Natalia Saiapova (MSc candidate). 2017. High-performance photometric stereo on the GPU.
  • Oleksandra Ganus (MSc candidate). 2016-2017. An RGBD system for acquiring full 3D-shape and reflectance of the human body.
  • Andreas Seibold (MSc candidate). 2016-2017. Single-view 3D-reconstruction using ballooning and shading. 50% supervision, with T. Mollenhof.
  • Christian Buttner (MSc candidate). 2016-2017. SVBRDF capture for quasi-planar surfaces
  • Hesam Rabeti (MSc candidate). 2016-2017. Real-time exposure correction on mobile devices.

Past:

  • Louis Fauvarque (MSc candidate). Summer 2016. Reading 3D-sonagrams on a smartphone. 60% supervision, with J.-D. Durou and G. Azzaro. Louis is now finalizing his MSc in INP-ENSEEIHT.
  • Jade Boumaza (MSc candidate). Summer 2015. Single view high-precision 3D-reconstruction. 60% supervision, with J.-D. Durou and F. Lauze. Jade is now a computer science engineer at Diginext.
  • Tom Lucas (MSc candidate). Summer 2015. Variational fusion of RGB-D data using color photometric stereo. 60% supervision, with J.-D. Durou and F. Lauze. Tom is now a software engineer at Thales.
  • Mathieu Pizenberg (MSc candidate). Summer 2014. Geometric and photometric calibration of a dermoscope. 60% supervision, with J.-D. Durou. Mathieu is now a Ph.D candidate at INP-Toulouse.
  • Richard Modrzejewski (MSc candidate). Summer 2014. Turning a mobile phone into a 3D-scanner. 60% supervision, with J.-D. Durou and P. Gurdjos. Richard is now a Ph.D candidate at ISIT (Univ. Clermont-Ferrand).

Scientific responsabilities

  • Member of the technical committee of the QCAV 2017 conference and of the SSVM 2017 conference
  • Co-organizer (with J.-D. Durou) of the workshop Photometric Stereo: From Theory to Industrial Applications, associated with the QCAV 2015 conference.
  • Reviewer for IJCV (Springer), JMIV (Springer), IEEE Trans. Image Processing, Siam Journal on Imaging Sciences, IET Computer vision, Computer and Graphics (Elsevier), and several conferences (ICCV, WACV, QCAV, SSVM, RFIA).