A scalable and composable analytics platform for distributed wide-area tracking

Published in 20th International Conference on Distributed Computing and Networking, 2019

As smart city deployments see an increase in the number of cameras available for surveillance, running complex analytics assisted by deep neural networks will become imperative if any meaningful insight is to be drawn from the feeds. While, neural networks have already surpassed human capabilities in certain tasks making them practical for real world deployments, there does not exist any platform that can support a wide-area deployment of these networks. We design and develop an analytics platform that allows easy composability of these neural networks to solve the problem of wide-area tracking. The platform solves multiple system side challenges that would otherwise hamper or complicate the development of such applications. It will also help in reducing the time and effort required for the development of such applications by exposing well defined APIs.