Baden-Baden 2011 (5-9 June), Germany

http://www.mrt.uni-karlsruhe.de/iv2011/

Workshop date: Sunday, June 5, full day

Organizer:   C. Curio, Max Planck Institute for Biological Cybernetics, Tübingen, Germany

Current driver assistance systems support the visual recognition of pedestrians in dynamic visual scenes, typically based on the detection of static and dynamic cues of individual persons. The integration of knowledge about more complex interactions between groups of humans into such systems is still a challenge but is necessary to enhance perception and decision making processes. This workshop presents interdisciplinary research and developments on human collective behavior analysis and synthesis, ranging from motion-capturing, graphical simulation for prototyping and experiments, machine-vision for detection and prediction, to robotics for safe navigation in crowded urban areas. Novel approaches at the interface of different disciplines with similar aims are contributing to a better understanding of human crowd behavior with the hope to ultimately lead to even more purposeful human-centered assistance systems. At the core of these approaches are generative models learned from real-world data, applied to the simulation of populated scenes and machine vision. The goal of this workshop is to give an overview and stimulate research and developments in this field ultimately leading to assistance systems that can take complex behavior of pedestrians in street scenes into account.

Topics: Human Crowd Perception, Environment Simulation, Pedestrian Detection and Tracking, Motion Capture, Models of Social Interaction, Intention Modelling, Dynamic Models, Social Robotics, Animation, Virtual Reality, Prototyping.

Julien Pettre, INRIA-Rennes, France, Martin A. Giese, Computational Sensomotorics, University Clinic, Tübingen, Germany, Luc van Gool, Computer Vision Laboratory, ETH Zürich, Switzerland, Kai O. Arras, Social Robotics Laboratory, Freiburg, Germany, Bernt Schiele, Max Planck Institute Informatics, Saarbrücken, Germany, Dariu Gavrila, Department of Environment Perception, Daimler Research & Development, Ulm, Germany, Christian Laugier, INRIA-Rhone-Alpes, Grenoble, France.

Crowd Simulation and Movement Representation

9:00-9:45 Dr. Julien Pettre

INRIA-Rennes/France

“Crowd Simulation for Interactive Virtual Environments”

Crowd simulation is a difficult problem. It is even more challenging in the context of interactive applications where computation time is strictly limited. Trade-offs between quality and performance are required to fit a simulation to interactive contexts. We discuss the various approaches to tune such trade-off and evaluate them.

10:00-10:45 Prof. Dr. Martin A. Giese

Computational Sensomotorics, University Clinic, Tübingen/Germany

“Synthesis of Body Movements Based on Learned Dynamic Primitives”

We present a method for the learning of generative models of individual and collective human behavior based on motion capture, exploiting a combination of unsupervised learning and nonlinear dynamics. Specifically, we show how stability constraints for coordinated collective behavior of groups of agents can be derived for such systems.

Social Interaction Models for Detection, Tracking and Prediction

11:00-11:45 Prof. Dr. Luc van Gool

Computer Vision Laboratory, ETH Zürich/Switzerland

“Crowd sorcering: trying to understand crowd flows”

We concisely discuss our work to track people in crowds. One possible improvement is the use of stronger models for their expected motions. We focus on a model that takes account of social influences when people move in crowds.

12:00-12:45 Prof. Dr. Kai O. Arras

Social Robotics Laboratory, Freiburg/Germany

“Range-Based People Detection and Tracking for Socially Aware Intelligent Vehicles”

We will present our line of work in detection and tracking of people and groups of people in 2D and 3D range data that combine tracking with learned or a priori given models of human activities and social behavior. This combination is shown to lead to improved tracking accuracy. We will also present very recent advances in people detection in dense 3D data.

13:00-14:15 Lunch break

Environment Perception and Path Planning

14:15-15:00 Prof. Dr. Bernt Schiele

Max Planck Institute Informatics, Saarbrücken/Germany

“Articulated Body Models and 3D Scene Models for Detection, Tracking and Prediction”
Work on articulated models has let to better detection and tracking results. Including 3D scene models can further improve performance but also enable better “understanding” of the environment.

15:15-16:00 Prof. Dr. Dariu Gavrila

Dept. of Environment Perception, Daimler Research and Development, Ulm/Germany

“From Detecting to Understanding Pedestrians”

General overview and Daimler activities.

16:15-17:00  Prof. Dr. Christian Laugier

INRIA-Rhone-Alpes, Grenoble/France

“Collision Risk Assessment and Application to a Risk based Navigation System”

This talk covers aspects of modeling mobile obstacle behaviors (vehicles, pedestrians), motion prediction for potential mobile obstacles, probabilistic collision risk assessment, risk based navigation decisions using incomplete task models and perception.

Panel discussion

17:15 - 17:30

TANGO