Generating New Ground Truth Data by Editing Previous Data from Integrated Video Annotation Database

Abstract

In order to develop object and event detection algorithms for intelligent CCTVs, a large amount of ground truth data is required, and it takes a lot of time and cost to generate the data. To produce large amount of ground truth data, an efficient video annotation system which can support re-creations of new video annotation data by retrieving and editing the previously stored ones from the integrated video annotation database is required. We developed a system that can retrieve and edit the previous annotation data to generate new ground truth data very efficiently. The newly generated annotation data can be used as the ground truth data for developing detection algorithms of intelligent CCTV. By using the developed video annotation system, a substantial amount of ground truth data which is needed for developing object and event detection algorithms can be generated easily in short time periods.