Recent technology has turned the video into an easily created
and widely spread media available as digital TV broadcasts,
Internet videos (YouTube, Video Google, BBC Motion Gallery),
professional digital archives and personal video collections.
This large and ever growing amount of data currently presents
a challenge for many video applications including video search,
browsing, content filtering and others. Human actions represent
the large and essential part of video content, hence, their
automatic interpretation is crucial to enable future video
applications.

This PhD work will address the challenge of action recognition
in realistic video. It will aim to overcome limitations of
current methods which are still far from addressing the full
range of video variability such as the individual variations of
people in expression, posture, motion and clothing; perspective
effects and camera motions; illumination variations; occlusions
and variation in scene surroundings. It will build upon and will
extend our recent work in this direction (see e.g.
http://www.irisa.fr/vista/actions ) using the modern tools of
computer vision, machine learning and multi-modal data mining.
The proposed work has potential of making high scientific impact
as well as to contribute to practical applications.

Scientific environment:
The successful candidate will work at INRIA centre in Rennes
under supervision of Ivan Laptev and Patrick Perez. The work
will be carried out within QUAERO programme http://www.quaero.org/

Requirements:
- MSc. degree (or a French Master Degree) in computer science,
  electrical engineering or related disciplines. Engineering
  degrees will be also considered for candidates with initial
  research experience.
- Background in one or several following subjects: computer
  vision, machine learning, image/video processing
- Good programming skills (preferably Matlab and C/C++)
- Good skills in written and spoken English

Salary: 1,537E Net per month

To apply, please send a full CV, a cover letter and contact details
of two referees to: Ivan Laptev (ivan.laptev at inria.fr)