REAL-TIME COMPUTER VISION IN CLOUDS THROUGH EFFECTIVE MONITORING AND WORKFLOW MANAGEMENT

Computer vision enables amongst others detection and tracking of static and moving objects, as well as identification of events and actions. Nevertheless the applicability and adoption of computer vision approaches in large-scale industrial environments is limited mainly due to their computation requirements when focusing on real-time objects tracking or events identification. In this paper we present the experimentation outcomes of a computer vision application that has been deployed on a large-scale multi-cloud facility. Effective monitoring and workflow management mechanisms are also presented as the enablers for meeting the real-time requirements of the computer vision application. We evaluate the effectiveness of these mechanisms through a set of experiments that demonstrate their value for allowing cloud infrastructures to provide real-time guarantees.