Automatic video annotation of human activities

• Grant number: 06/06377-5
• Support Opportunities: Scholarships abroad - Research
• Start date: March 16, 2007
• End date: August 15, 2007
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: Carlos Hitoshi Morimoto
• Grantee: Carlos Hitoshi Morimoto
• Host Investigator: Rama Chellappa
• Host Institution: Instituto de Matemática e Estatística (IME). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Institution abroad: University of Maryland, College Park, United States

Abstract

The explosion of the amount of multimedia information that is available to everyone has created the need for more efficient solutions to manage and search visual information in large video databases. Nowadays, the annotation of visual information is a tedious manual task, and therefore the resulting annotated video is subject to errors and inaccuracies. There is clearly the need to develop automatic tools for video annotation. During the 5 months period of my visit, we will explore new alternatives to object image segmentation in video, in particular to detect humans and identify poses and some activities. Example of scenes that we might be able to annotate automatically are dialog scenes, people walking, running, and carrying objects. This project is related to the FAPESP project 04/15717-9, still active and under my coordination, and will be developed at the Center for Automation Research (CfAR), with the collaboration of Prof. Rama Chellappa. (AU)