By Wael Guibene & Hossam Hmimy, Charter Communications
A defining moment of the century, so far, is the unprecedented impact that COVID-19 has brought to the economies world-wide, the populations and defining new norms in society.
Our paper details how we can enforce the new rules of society like social distancing and wearing face masks in open-spaces using computer vision and deep learning through:
- Detecting people on a particular scene,
- Calculating and monitoring the distances between the different people,
- Tracking movements and segregating moving people (might come close to each other during brief moments) from people standing still and violating the social distancing rules, and
- Creating alerts (audio, visual, light…) to enforce social distancing.
Our approach also ensures that, in confined spaces, we count people and create visual and audio alerts when the number of people exceeds the Center for Disease Control (CDC) guidelines (10 people per room).
We detail in the paper the computer vision and deep learning frameworks we used to achieve high confidence in detecting human presence, calculating and calibrating distances in the frames, and removing false positives (eg. people crossing paths while walking versus people standing still).