Pierre Zins

Pierre Zins

Computer Vision / Deep Learning Research Engineer (PhD)

Inria Grenoble - Rhône-Alpes

Biography

I am a Computer Vision Research Engineer. In 2023 I received my PhD from University Grenoble Alpes for my work on “3D shape reconstruction from multiple views”. My work has been published in top international conferences (3DV and CVPR).

As PhD student I worked in the Morpheo team at Inria Grenoble Rhône-Alpes. My thesis was part of a collaboration between Meta Reality Labs in San Francisco and Inria . I was supervised by Stefanie Wuhrer and Edmond Boyer (Inria Morpheo) and Tony Tung (Meta Reality Labs).

My research focused on 3D shape reconstruction from images and was divided along three axes:

  • 3D reconstruction of dressed humans from a few sparse views by leveraging a neural implicit representation and attention mechanisms.
  • 3D reconstruction of shapes from multiple views using an SRDF (Signed Ray Distance Functions) representation.
  • Improved implicit shape modeling from a few views by using multi-view constraints

Before starting my PhD, I have worked for 1 year at Wrnch, as a computer vision engineer. In 2018 I graduated from Polytechnique Montréal (Master of Applied Science in Computer Science) and Université de Technologie de Compiègne (French Engineer degree in Computer Science). I did my Master Thesis at the Distributed Open Reliable Systems Analysis Lab from Polytechnique and work on performance analysis of machine learning dataflow applications executing in heterogeneous environments. I have also spent 6 months as an exchange student at Technische Universität Graz in Austria.

Download my resume.

Interests
  • Computer Vision
  • 3D reconstruction
  • Multi-View Geometry
  • Deep Learning
  • Computer Graphics
  • Algorithms
Education
  • PhD in Computer Vision, 2019-2023

    Inria / Université Grenoble Alpes

  • MSc in Computer Science, 2017-2018

    Polytechnique Montréal

  • Engineer degree in Computer Science, 2012-2018

    Université de Technologie de Compiègne

Experience

 
 
 
 
 
PhD Student in Computer Vision
Oct 2019 – Apr 2023 Grenoble, France

During my thesis, I have been working on 3D shape reconstruction from images. In particular, our work is divided along three axes:

  • 3D reconstruction of dressed humans from a few sparse views by leveraging a neural implicit representation and attention mechanisms.
  • 3D reconstruction of shapes from multiple views using an SRDF (Signed Ray Distance Functions) representation.
  • Improved implicit shape modeling from a few views by using multi-view constraints.
 
 
 
 
 
Computer Vision Engineer
Oct 2018 – May 2019 Montreal, Canada
I worked on several computer vision applications that use human pose estimation and on a multi-camera platform for 3D markerless motion capture (multi cameras calibration, video feed synchronization, 3D estimation from several 2D poses).
 
 
 
 
 
Master Thesis
Jan 2017 – Aug 2018 Montreal, Canada
I developped performance analysis tools for machine learning dataflow applications executing in heterogeneous environments. I focused on the library TensorFlow and its dataflow computation graph. The goal was to develop tools which will help to understand the performance of the applications and to detect limiting elements or bottlenecks. A main aspect was to insure that the available hardware (CPUs and GPUs) is used efficiently.
 
 
 
 
 
Intern
Sep 2015 – Feb 2016 Tourcoing, France

I worked on different projects:

  • Continuous Integration System Buildbot, Docker, Wakeonlan, scripts Bash.
  • C++ Dévelopment for a license system C++, CMake, Boost, XML, Client-Server.
  • Smart C++ tools for memory allocations tagging and monitoring.