Section: New Results

Introduction

This year Pulsar has tackled several issues related to its two main research axes : scene understanding for activity recognition and software engineering for activity recognition.

Scene Understanding for Activity Recognition

Participants : Slawomir Bak, Piotr Bilinski, Bernard Boulay, François Brémond, Guillaume Charpiat, Duc Phu Chau, Etienne Corvée, Julien Gueytat, Ratnesh Kumar, Vincent Martin, Sabine Moisan, Emmanuel Mulin, Jose-Luis Patino Vilchis, Guido-Tomas Pusiol, Leonardo Rocha, Rim Romdhame, Silviu Serban, Malik Souded, Monique Thonnat, Sofia Zaidenberg, Daniel Zullo.

This year Pulsar has proposed new algorithms in computer vision (people head and face detection and people re-identification), in reasoning (activity recognition and uncertainty handling). More precisely, the new results for this research axis concern:

• People detection in monocular video sequences ( 6.2 )

• Online Parameter Tuning for Object Tracking Algorithms ( 6.3 )

• Fiber Based Video Segmentation ( 6.4 )

• Multiple Birth and Cut Algorithm for Multiple Object Detection ( 6.5 )

• Exhaustive Family of Energies Minimizable Exactly by a Graph Cut ( 6.6 )

• Steepest Descent in Banach Spaces with Application to Piecewise-Rigid Evolution of Curves ( 6.7 )

• Object Tracking Using a Particle Filter based on SIFT Features ( 6.8 )

• Human Re-identification using Riemannian Manifolds ( 6.9 )

• Global Tracking of Multiples Actors ( 6.10 )

• Crowd Data Collection from Video Recordings ( 6.11 )

• Events Recognition and Performance Evaluation ( 6.12 )

• Group interaction and group tracking for video-surveillance in underground railway stations ( 6.13 )

• Action Recognition in Videos ( 6.14 )

• Activity Recognition Applied on Health Care Application ( 6.15 )

• A Cognitive Vision System for Nuclear Fusion Device Monitoring ( 6.16 )

• Scenario Recognition with depth camera ( 6.17 )

• Trajectory Clustering for Activity Learning ( 6.18 )

Software Engineering for Activity Recognition

Participants : François Brémond, Bernard Boulay, Hervé Falciani, Daniel Gaffé, Julien Gueytat, Sabine Moisan, Annie Ressouche, Jean-Paul Rigault, Leonardo Rocha, Sagar Sen, Daniel Zullo.

This year Pulsar has improved the SUP platform. This latter is the backbone of the team experiments to implement the new algorithms proposed by the team in perception, understanding and learning. We improve our meta-modeling approach to support the development of video surveillance applications based on SUP. We continue the development of a scenario recognition module relying on formal methods to support activity recognition in SUP platform. We also continue to study the definition of multiple services for device adaptive platform for scenario recognition. Finally, we are implementing the new theoretical results obtained last year to improve the Clem toolkit.

The new results related to this research axis concern:

• SUP Software Platform ( 6.19 )

• Model-Driven Engineering and Video-surveillance ( 6.20 )

• Scenario Analysis Module ( 6.22 )

• Multiple Services for Device Adaptive Platform for Scenario Recognition ( 6.23 )

• The Clem Toolkit ( 6.24 )