Non-Rigid Object Tracker Based On a Robust Combination of Parametric Active Contour and Point Distribution Model

in Proceedings of the SPIE Conference on Visual Communication and Image Processing (2007)

Sequential Variational Inference for Distributed Multi-Sensor Tracking and Fusion

in The 10th International Conference on Information Fusion (2007)

Approximate Policy Iteration for Closed-Loop Learning of Visual Tasks

(2006, September)

Approximate Policy Iteration (API) is a reinforcement learning paradigm that is able to solve high- dimensional, continuous control problems. We propose to exploit API for the closed-loop learning of ... [more ▼]

Approximate Policy Iteration (API) is a reinforcement learning paradigm that is able to solve high- dimensional, continuous control problems. We propose to exploit API for the closed-loop learning of mappings from images to actions. This approach requires a family of function approximators that maps visual percepts to a real-valued function. For this purpose, we use Regression Extra-Trees, a fast, yet accurate and versatile machine learning algorithm. The inputs of the Extra-Trees consist of a set of visual features that digest the informative patterns in the visual signal. We also show how to parallelize the Extra-Tree learning process to further reduce the computational expense, which is often essential in visual tasks. Experimental results on real-world images are given that indicate that the combination of API with Extra-Trees is a promising framework for the interactive learning of visual tasks. [less ▲]

Task-Driven Discretization of the Joint Space of Visual Percepts and Continuous Actions

in European Conference on Machine Learning (2006)

Multi-view object tracking using sequential belief propagation

in Computer Vision – ACCV 2006 (2006)

Multiple cameras and collaboration between them make possible the integration of information available from multiple views and reduce the uncertainty due to occlusions. This paper presents a novel method ... [more ▼]

Multiple cameras and collaboration between them make possible the integration of information available from multiple views and reduce the uncertainty due to occlusions. This paper presents a novel method for integrating and tracking multi-view observations using bidirectional belief propagation. The method is based on a fully connected graphical model where target states at different views are represented as different but correlated random variables, and image observations at a given view are only associated with the target states at the same view. The tracking processes at different views collaborate with each other by exchanging information using a message passing scheme, which largely avoids propagating wrong information. An efficient sequential belief propagation algorithm is adopted to perform the collaboration and to infer the multi-view target states. We demonstrate the effectiveness of our method on video-surveillance sequences. [less ▲]

Data Fusion by Belief Propagation for Multi-Camera Tracking

in The 9th International Conference on Information Fusion (2006)

Robust Non-Rigid Object Tracking Using Point Distribution Manifolds

in 28th Annual Symposium of the German Association for Pattern Recognition (DAGM) (2006)

Object Tracking Using Color Interest Points

Conference (2005)

Learning, then Compacting Visual Policies

in 7th European Workshop on Reinforcement Learning (2005)

Task-Driven Learning of Spatial Combinations of Visual Features

in Proc. of the IEEE Workshop on Learning in Computer Vision and Pattern Recognition (2005)