Automatic monitoring of social distance via CCTV

posted by Abishek Muthian, permalink  

Social Distancing or more accurately Physical Distancing is the key prevent the spread of COVID-19 before a vaccine is made available for everyone on the planet.


But since lockdowns cannot continue indefinitely due to economic reasons, countries which were in lockdown are opening up. This brings a new challenge of how to monitor and enforce social distancing when everyone is back on the roads.

CCTVs would be crucial in monitoring and enforcing social distancing in public places and in workplaces; The manual overhead in doing it is very high, fortunately if there's one thing which AI has proven itself as more capable than human beings it's computer vision.

But, since average CCTVs don't have depth perception, have different views and people are not in the same plane of axis; judging the distance between two people from video footage is a hard problem. Fortunately, Andrew Ng's company has released a tool to do just this and has released how the procedure works - https://landing.ai/landing-ai-creates-an-ai-tool-to-help-customers-monitor-social-distancing-in-the-workplace/.

It solves the problem of calculating distance by using homography to get top-down view of the image from video, using a manual calibration tool. Unfortunately, none of this has been open-sourced as of writing and hence a need gap for this to be released as an open-source tool for any govt. or workplace to make use of.

Such a tool can also be used on open CCTV feeds from Shodan, Insecam to analyse social distancing in countries around the world.
Register Login to comment or vote on this problem
Voted!
Need karma! Please check contributor guidelines.