Towards Automating Social Distance Violations using AIoT

The pandemic caused by the COVID 19 virus had a tremendous impact on the retail and tourism industry resulting in a loss of around two trillion dollars per year. Recovering completely from the pandemic may take several years and safety guidelines like social distancing will remain a norm for many years to come. Many governments have levied heavy fines for businesses that fail to implement social distancing guidelines. Implementing manual social distancing policies can be expensive and error prone. This paper proposes a smart system that uses cameras along with Artificial Intelligence Internet of Things (AIoT) edge devices like the Nvidia Nano GPU to automatically detect social distance violations on the edge. The system achieved an accuracy of 92.2% and an F1-Score of 0.851 using a single camera only. On an Nvidia Nano, the system has a very acceptable rate of 1.7 frames per second on the edge obviating the need for any expensive back-end transmission and processing. The system uses the MQTT protocol to instantly notify the manager or owner of the business via a mobile App and a Web application for necessary action.