Toward robust analysis of unstructured point-cloud

I am a CNRS researcher, working at the STORM research group, in the IRIT laboratory, Toulouse.
I am also an associate researcher of Archeovision, a CNRS laboratory studying how Cultural Heritage researches can be leveraged by the acquisition, modeling and processing of 3D data.

My research interests focus on the analysis of unstructured data (usually point-clouds) generated by acquiring a physical object or scene, a very hot topic in several research communities (Computer&Graphics, Pattern Recognition, …) and in the industry. Above all the research aspects, I’m considering the following challenges: 1) there is currently no practical formalism allowing to characterize shapes at locals, medium and large scales at once, 2) the notion of a shape scale within an object is still ill-defined, despite a wide range of attempts, 3) very few studies have been published on the transitions between the pertinent scales and shapes of a 3D model.

In addition to these theoretical challenges, I’m also very interested in the applications that could be enhanced by more efficient analysis of raw point clouds: cleaning, reconstruction, abstraction, completion, recognition, etc.

I’ve also a growing interest on color processing, using geometrical properties of colors. We call that Structural Color Processing.


We are recruiting students and post-doc,
Checkout details and other open calls on our group webpage. Spontaneous application are also welcome: contact me by email with a project abstract or proposal.

News

  • December 11th 2019: Our paper Persistence Analysis of Multi-scale Planar Structure Graph in Point Clouds has been conditionally accepted at Eurographics 2020. More details coming soon.
  • December 3rd 2019: Today I was invited to participate to the round-table discussion on the topic: “From the archeological studies to the 3D models: challenges and perspective on acquisition and interaction”, organized by the Consortium 3D-SHS. Other participants: Valérie Gouranton, Ronan Gaugne, Valentin Grimaud. [tweet].
  • November 22th 2019: Anahid Ghazanfarpour has successfully defended her PhD today, entitled: Proximity-Aware Multiple Meshes Decimation using Quadric Error Metric.
  • November 5th 2019: The CNRS is communicating on our paper: L’imagerie lidar simple-photon passe en temps réel (in french)
  • November 2d 2019: Our paper Real-time 3D reconstruction from single-photon lidar data using plug-and-play point cloud denoisers is now published in Nature Communications and available online.
  • September 23th 2019: Our project Making a 3D dataset for Cultural Heritage studies in Computer Sciences has been accepted by the french Consortium 3D SHS.
  • September 17th 2019: Our paper Real-time 3D reconstruction from single-photon lidar data using plug-and-play point cloud denoisers has been conditionally accepted at Nature Communications.
  • May 21th 2019: New paper under submission and published on arXiv, entitled Real-Time 3D Reconstruction of Complex Scenes Using Single-Photon lidar: When Image Processing Meets Computer Graphics.
  • May 18th 2019: Google Summer of Code 2019 has started, Necip Yildiran is working on making OpenGR working directly on CGAL data-structures.
  • April 15th 2019: New paper accepted at Computer & Graphics, first authored by Sara Casti, and entitled Skeleton Based Cage Generation Guided by Harmonic Fields.
  • April 10th 2019: Our contribution to the SHREC’19 track Feature Curve Extraction on Triangle Meshes will be presented at the 2019 Eurographics Workshop on 3D Object Retrieval.
  • February 27th 2019: I am co-mentoring a project for the next Google Summer of Code with CGAL. Links:
  • previous entries …

Affiliations:

  • CNRS, IRIT, Université de Toulouse, INPT, UPS, UT1C, UT2J, France.
  • CNRS, Archeovision, Université de Bordeaux, France.

Timeline

  • 2016-Now: CNRS researcher (chargé de recherche) at IRIT (Toulouse, France), at the STORM research group.
  • 2014-2016: Post-doc at IRIT, Université Paul-Sabatier (Toulouse, France), with David Vanderhaeghe,
  • 2013-2014: Post-doc at University College London working with Niloy Mitra,
  • 2012-2013: Post-doc at Inria Bordeaux Sud-Ouest (France), working with Pascal Barla, Gaël Guennebaud and ArcheoVision,
  • 2009-2012: PhD thesis at Inria Bordeaux Sud-Ouest (France), supervised by Patrick Reuter and Christophe Schlick.

Other activities

  • 2019-Now: Scientific consultant for ZEG.ai.

Research interests (see publication and research pages for details)

  • 3D shape analysis: focusing on acquired data, possibly with complex geometry and multiple pertinent scales
  • 3D shape registration: focusing on global (no input pose) rigid registration with partial overlap,
  • Semi-automatic systems: mixing efficient computations and interaction for real-time solutions,
  • Structural Color Processing.

Students

Opportunities: checkout https://www.irit.fr/STORM/site/open-positions/.

PhD students:

  • Jie Shao (co-advisor: Wuming Zhang): point-cloud registration and processing
  • Thibault Lejemble (co-advisor: Loic Barthe): shape analysis
  • Anahid Ghazanfarpour (co-advisor: Jean-Pierre Jessel): mesh processing

Engineers (Scientific Supervision):

  • Clément Rodrigues Viguier (co-supervisor: David Vanderhaeghe) [Méso-Star]: Structural Color Processing
  • Chems-Eddine Himeur (co-supervisor: Mathias Paulin): Appearance-preserving mesh decimation

Master students:

Undergrad students:

Former students:

  • Chems-Eddine Himeur: Deep Learning for point-based shape analysis
  • Filippo Fanni (co-advisor: David Vanderhaeghe): Structural Color Processing
  • Paul Bernardi (co-advisor: David Vanderhaeghe): Structural Color Processing
  • Céline Michaud (co-advisor: Mathias Paulin. Company: Oktal-SE): mesh processing
  • Baptiste Delos (co-advisor: David Vanderhaeghe): Structural Color Processing
  • Fatemi Arezou (co-advisor: David Vanderhaeghe): real-time rendering
  • Xavier Chalut (co-advisor: Jean-Pierre Jessel): mesh processing
  • Georges Nader (co-advisor: Gaël Guennebaud): shape analysis
  • Bastien Perpère (co-advisors: Gaël Guennebaud, Pascal Barla. Company: The Foundry): rendering