Social Distancing Using AI For Health/Safety (COVID-19)

Challenges.

Different approaches tried to ensure social distances like using ground markers; Bluetooth enabled devices, which give haptic feedback if it finds a nearby device too close, and even manual monitoring. Many times, it is not possible to enforce a strict use of such mobile devices or deploy staff to monitor this.

Solution.

Monitoring distance between people is a computer vision problem and can be done 24x7 with minimal manual intervention.  
Since CCTV camera’s are mostly installed in corners and they capture 2D frames in videos, images cannot be directly use for estimating distances. It is necessary to use Computer Vision / AI techniques to detect people’s position and distance in a virtual Bird’s eye view. 
We have developed an accelerator – “Inspection AI” which using CCTV cameras can estimate a bird’s eye view of area in real time and can predict the distance between detected people in inches/feet 
Solution requires an initial calibration step to tell the system about scene perspective and how to scale pixels to distance.  
With it’s real time AI backbone (our Vision AI accelerator for computer vision), solution can be extended to send real time alert/notification, generate floor density heat map, and enable to analyze trend of how well social distancing is maintained in the area

The Impact.

Keeping people healthy & safe in the times of a global pandemic (COVID-19).  Can be applied to factories and warehouses where people (using heavy/dangerous equipment's) must keep a safe distance.