A modular architecture for optimal video analytics deployment

Efficient video camera network design ensures performance of networked cameras in video surveillance systems, but also in virtual reality, scene reconstruction, etc. These applications require efficiency at different performance level. Several authors have addressed this problem in the past, but none considered the Human in the Loop (HIL). In this paper we propose a modular sequential architecture that includes interaction with an end-user in order to obtain a deployment solution in near-real time. A possible solution is outlined and their related tradeoffs are presented.