Hey, vitrivr! - A Multimodal UI for Video Retrieval

Prateek Goel, Ivan Giangreco, Luca Rossetto, Claudiu Tănase, Heiko Schuldt
In Proceedings
Appears in
Proceedings of the European Conference on Information Retrieval (ECIR)
Aberdeen, Scotland, UK

In this paper, we present a multimodal web-based user interface for the vitrivr system. vitrivr is a modern, open-source video retrieval system for searching in large collections of video using a great variety of query modes, including query-by-sketch, query-by-example and query-by-motion. With the multimodal user interface, prospective users benefi t from being able to naturally interact with the vitrivr system by using spoken commands and also by applying multimodal commands which combine spoken instructions with manual pointing. While the main strength of the UI is the seamless combination of speech-based and sketch-based interaction for multimedia similarity search, the speech modality has shown to be very eff ective for retrieval on its own. In particular, it helps overcoming accessibility boundaries and o ffering retrieval functionality for users with disabilities. Finally, for a holistic natural experience with the vitrivr system, we have integrated a speech synthesis engine that returns spoken answers to the user.