Performance Evaluation of an IoT Edge-Based Computer Vision Scheme for Agglomerations Detection Covid-19 Scenarios

Edge architectures have emerged as a solution for the development of Internet of Things (IoT) applications, especially in scenarios with ultra-low system latency requirements and a huge amount of data transmitted in the network. This architecture aims to decentralize systems and extend cloud resources to devices located at the edge of networks. Various benefits regarding local processing, lower latency, and better communication bandwidth can be highlighted. This study proposes an edge architecture which uses computer vision to detect people in agglomerations. To evaluate the performance of the proposed architecture, an use-case for agglomeration detection in the Covid-19 scenarios is presented. A comparative analysis of the detection is performed through videos from a public database. The obtained results demonstrate a gain in terms of computational performance with a video analysis in comparison to the best solutions available in the literature. The proposed solution can be a powerful edge tool to support the combat against Covid-19 Pandemic.