DANTE: A Video Based Annotation Tool for Smart Environments

This paper presents a novel system, which uses a set of video cameras to support the manual annotation of object interaction and person to object interaction within smart environments. The DANTE (Dynamic ANnotation system for smart Environments) system uses two stereo-based cameras to monitor and track objects, which are tagged with inexpensive paper based fiducial markers. Offline, the DANTE annotation module allows users to navigate, frame-by-frame, through recorded sessions and to annotate interesting and relevant events. The generation of annotated datasets has significant utility within the domain of smart environments to support the development of data driven context-aware systems which require large training datasets. The current paper presents the rationale for this work, provides a technical description of the system and highlights its potential opportunities. Early experimental work, investigating a comparison with accelerometer based sensors is also briefly introduced.