18-month postdoctoral position in Computer Vision at GIPSA-Lab,
Automatic annotation of skilift videos
Work place: Grenoble campus, France
Research Laboratory: GIPSA-Lab, Grenoble
Hiring institution: Grenoble Institute of Technology
The MIVAO project gathers several academic and industrial partners and aims
at building a new system mixing computer vision and machine learning. The
leading partner Bluecime, a start-up company from Grenoble (France),
develops computer vision solutions to ensure user safety on
skilifts. Its main product is already running in four French resorts and
At the moment, the system is able to detect and track each seat (aka
to know if there are people or not on it and to determine the position
(high or low) of the safety railing.
These informations allow to specify whether a situation is dangerous to
trigger an alarm. Nowadays, the system reaches an accuracy about a
dangerous situation of 98,47% (value acquired by tests on a database
made of about 25,000 vehicle tracks).
Although the results are satisfactory, the algorithm has to take a lot
of variabilities into account (weather situation, illumination, topology
of skilift, type of vehicle), that prevent both an adaptive install and
the development of extra functionalities. In this project, these two
will be tackled.
The postdoc candidate will have in charge the development of innovative
vision / machine learning techniques to perform automatic annotation
(taking into account all previously cited variabilities) of
the videos that will describe the scene : number of passengers and their
location, presence of children, weather situation, season, …
The candidate will investigate two different research fields, one about
dimensionality reduction and the second about 3D modeling (of the seat).
? PhD in computer science,
? Experience in C / C++ or Python,
? Knowledge in nonlinear dimensionality reduction or 3D modeling
constitutes an advantage.