08/04/2021
The Human Resources Strategy for Researchers

Phd contract in the field of Engineering financed for three years by the University of Clermont Auvergne

This job offer has expired


  • ORGANISATION/COMPANY
    Université Clermont Auvergne
  • RESEARCH FIELD
    Engineering
  • RESEARCHER PROFILE
    First Stage Researcher (R1)
  • APPLICATION DEADLINE
    10/06/2021 00:00 - Europe/Brussels
  • LOCATION
    France › Aubière
  • TYPE OF CONTRACT
    Temporary
  • JOB STATUS
    Full-time
  • HOURS PER WEEK
    35
  • OFFER STARTING DATE
    01/10/2021
  • REFERENCE NUMBER
    SPI-CD-2021-007

OFFER DESCRIPTION

Semi-supervised learning for 3D reconstruction of environmentsfrom images

The 3D reconstruction of an environment from images is useful in a lot of applicationsincluding virtual reality and autonomous vehicles.Several methods of computer vision and photogrammetry are needed to solve this problemin three steps: the estimation of point clouds and camera parameters (6DoF motion andintrinsic ones), the surface reconstruction, and the texturing.The ComSee team in the Institut Pascal has contributions in all these topics.One of the most promising ways of improvements is the use of deep learning (DL) methodsto solve the surface reconstruction step.An objective is also to avoid the supervised methods, that need databases with 3Dinformation recovered by a scanner.There are several reasons to do this: price/availability of the scanner and time/ease ofacquisition.Here we propose to use a database composed of several large environments that arereconstructed by a previous non-DL method, with a minority of manual correctionsof thereconstructed surface.One idea is that the network learns to replace a piece of surface, that can be bad, by a goodone.Since this looks like a projection (an idempotent function), a possible choice of network is anauto-encoder.Furthermore, thesize of the local area of the network computation is a tradeoff: small enoughfor efficiency, large enough so that the network can recover semantic information and decidehow to correct the surface.Thanks to learning, we expect to improve the result of asurface reconstruction method (thatis not DL), eg if the experimental conditions are more difficult.There are many kind of environments and surface reconstruction methods that can be usedto generate a database.We mostly focus on the case of outdoor environments (both urban and natural) reconstructedby using a helmet-held consumer-grade 360 camera, but other cases can be studied.This case is interesting for both applications (eg virtual reality) and ease of databasegeneration.

More Information

Offer Requirements

  • REQUIRED EDUCATION LEVEL
    Engineering: Master Degree or equivalent
Work location(s)
1 position(s) available at
The Pascal Institut (IP)
France
Région Auvergne Rhône-Alpes
Aubière
63178
Campus Universitaire des Cézeaux, TSA 60026, CS 60026, 4 Avenue Blaise Pascal

Open, Transparent, Merit based Recruitment procedures of Researchers (OTM-R)

Know more about it at Université Clermont Auvergne

Know more about OTM-R

EURAXESS offer ID: 625351

Disclaimer:

The responsibility for the jobs published on this website, including the job description, lies entirely with the publishing institutions. The application is handled uniquely by the employer, who is also fully responsible for the recruitment and selection processes.

 

Please contact support@euraxess.org if you wish to download all jobs in XML.