Post-doc positions

DYNAMIC REMOTE SENSING (Inria postdoc contract of 18 months (gross salary: 2500 euros per month))

  • Announcement date: January 19, 2009

  • Group: Vista group (http://www.irisa.fr/vista) and new Fluminance group at INRIA Rennes Center

  • Start: as soon as possible

  • Context: joint Microsoft/Inria Research Lab

  • Research program: Satellite images, whose spectral, spatial and temporal resolutions are constantly growing, offer a unique way to observe and study our environment and its evolutions. As opposed to in situ measurements with probes, they indeed give access, in a versatile way, to observations densely spread over the planet. With the increasing interest of scientific, economical and political spheres into critical and intertwined questions on climate changes, ecological catastrophes and resource management, the need for tools to process, analyse, annotate and search such massive image data becomes pressing.

    The general problem on which we want to focus concerns the detection, the analysis and the tracking of "dynamic events" of interest in satellite images. We aim in particular at studying the design of generic probabilistic learning tools, that include physically-grounded dynamical prior when available, for the following tasks:

    1. Detection of salient changes in multi-temporal satellite images. This problem is related to the classic problem of motion detection in videos but exhibits important specificities due to the very low cadences (from 15 minutes to several months between two consecutive images of the series under study).
    2. Detection of specific types of changes: similarly to the problem of detecting certain categories of objects in images, detectors of pre-defined types of dynamic events (e.g., eye of hurricane, fire start, convection of cloud cells) could be learnt from appropriately annotated data.
    3. Tracking high-dimensional entities: atmospheric or oceanic velocity fields, vorticity maps of these fields, complex shapes of atmospheric and oceanic structures, among others, are very high-dimensional dynamical entities. Probabilistic sequential estimation of such entities still defeats existing tracking tools.

    The emphasis among these tasks remains open and might depend on applicant's taste and background.

  • Requirements:
  • Contact: Please send detailed CV, list of publications, letter of intent and recommendation contacts to Patrick Perez (Patrick.Perez@inria.fr) and Etienne Memin (Etienne.Memin@inria.fr)

DETECTION, COUNTING, TRACKING AND MOTION ANALYSIS OF PEDESTRIANS IN VIDEOS (Inria postdoc contract of 12 months, renewable (gross salary: 2500 euros per month))

  • Announcement date: January 19, 2009

  • Group: Vista group (http://www.irisa.fr/vista) at INRIA Rennes Center

  • Start: as soon as possible

  • Research program: This works is part of "Cipebus" project with Inrets (http://www.inrets.fr/index.e.html), Versailles city council and several industrial partners (including Citilog, http://www.citilog.com/index_en.php). The overall goal of this project is the real-time monitoring of crowded street crossings with multiple cameras for commanding and testing of various traffic light management policies.

    In this project, our group will focus on the monitoring of people approaching and/or taking pedestrian crossings. More precisely, we aim at detecting them, tracking them as groups or individuals, analyzing their movements as well as the ones of surrounding vehicles. Based on the large expertise of our group in motion analysis (detection, segmentation, interpretation, tracking), we want to address the following challenges:

    • detection of people and moving objects in highly cluttered scenes
    • tracking of persons, groups of persons and vehicles
    • learning of motion patterns for each type of entity
    • classification of various types of dynamical situations
    • combination of previous algorithmic bricks with the ones already existing in Citilog systems.

  • Requirements:
    • Ph.D. in computer vision, image processing or machine learning.
    • good programming expertise (C/C++, Matlab, Visual, Windows, MFC, OpenCV, Linux)

  • Contact: Patrick Perez (perez@irisa.fr)


  Webmaster : vista_webmaster@irisa.fr
  Last modified: 19-01-2009