Active database interface for video search

Abstract

Surveillance video is abundant in urban areas and can often be vital to police when conducting investigations. Currently, the only means investigators have of finding a suspect in video is to watch it from beginning to end. This can be tedious and waste numerous man hours that could be expended elsewhere. Survq is a system that aids police detectives by automatically identifying key features of video using novel machine learning algorithms. These features are then used to filter video to those that match a suspect description. In this paper, we present the interface used for investigation and how it was designed with this use case and underlying system in mind.