August 1, 2020
TLDR: I built an (open source) app that automatically scolds anyone not wearing their COVID-19 mask. I’d love your help figuring out how to actually make it impactful in your local area.
I wish I could yell at each a**hole that isn’t wearing a mask around me. The burden gets even bigger for small business owners, teachers, office workers, bus drivers, and so on.
Then I saw this in an NYT newsletter:
Yeesh. So I thought… when else in history has humanity had issues adopting a simple, low-effort habit to save lives?
Seat belts!
And you know what worked better to increase seat belt usage than any education or law enforcement?
Annoying, automated ding ding ding sounds in cars.
Introducing the ???FaceMaskScolder3000! Let the machine do the scolding for you.
How to use: Load the site on a laptop or tablet* and point the webcam at the entrance to your space. The alarm will go off whenever someone walks in without a mask fully covering their mouth and nose.
You may optionally add your own judgmental stare.
There’s plenty of companies that provide this commercially. This is the weekend-warrior open-source version that can run on any device with a browser and a webcam.
[Ok, maker hat off, pragmatic hat on] Now the real bottlenecks here include: awareness, cost, distribution, implementation, maintenance, support, etc. This hack was the easy/sexy part. If you’ve read this far, perhaps you see a way to solve those challenges in your local neighborhood.
☝️☝️ I’d love to be in touch about this. ☝️☝️
*Most people don’t have an extra laptop/tablet laying around. (And if you do, you should donate it to families who could really use it for remote learning.) A friend suggested making a raspberry pi version to bring down costs… open to more cost reducing ideas!
Nerdy details:
This face detection web app runs entirely on the frontend, and does not store any videos or images, neither on your device nor on any servers.
You can view source code or fork your own using Glitch.
I originally forked off of a tutorial by CodeSTACKr (Github) which makes use of the incredible face-api.js. Special thanks to Dan Coldham for the elegant solution to a big false positives issue. I also found a way more data-sciency how-to using pytorch.
Learn about the wonders of masks from the NYTimes or UCSF.
The app, code, etc. are all free to use non-commercially (I mean, like a business can use it but not sell it, you get the spirit – maybe this isn’t the right license? Would love advice.)
Dedicated to Justin Hunter, a 17 year old who lost both parents to COVID-19 and who said: “If you don’t wear it for yourself, wear it for the next person, because you could be saving that person’s life.”