Für eine korrekte Darstellung dieser Seite benötigen Sie einen XHTML-standardkonformen Browser, der die Darstellung von CSS-Dateien zulässt.


Semantics of Human Behavior in Image Sequences

Prof. Dr. Jordi Gonzalez and Marco Pedersoli, CVC Barcelona

Date: 05/10/2011
Time: 09:00-10:40
Room: PB-H 0103 (ZESS Lecture Room)

Images, video, or multimedia are words that currently sound familiar to the majority of people. An enormous amount of video is daily produced by surveillance systems or broadcast companies, but also by travellers who want to keep the memories of new places visited. Considering such an amount of multimedia data, its analysis, processing, indexing and retrieval is a truly challenging task.

From this point of view, automatic image and video scene understanding is of importance. The main task of scene understanding is to give a semantic interpretation to observed images and video. In other words, scene understanding tries to bridge the semantic gap between the low-level representation of images and videos and the high-level, natural language description that a human would give about them. In our work, only those scenes including humans will be considered, as they are by far the most common ones in the studied domains. Nevertheless, emphasis will be on the interaction of these humans with their environment, since such a global approach will be proven to provide more information than separate analysis.

We will also present a method that can dramatically accelerate object detection with part based models. The method is based on the observation that the cost of detection is likely to be dominated by the cost of matching each part to the image, and not by the cost of computing the optimal configuration of the parts as commonly assumed. Therefore accelerating detection requires minimizing the number of part-to-image comparisons. To this end we will propose a multiple-resolutions hierarchical part based model and a corresponding coarse-to-fine inference procedure that recursively eliminates from the search space unpromising part placements.