In a real world, you actually want to do this when __anyone__ is passing by. You don't want to be seen as a slacker by your boss, but you also don't want to be seen as a slacker by your colleagues.
In which case, this is actually a much easier problem to solve since all you need to do is detect any motion (instead of all this convoluted--although hilarious--deep learning task).
This is like the story where some government tried to come up with a super high-tech pen that works in the space (because regular pens don't work in zero gravity environment) when the final answer was just a pencil.
So the lesson here is: there are many ways to solve a problem, and if you don't know what problem you're trying to solve and you don't know your customer well enough, you will end up wasting a lot of resource doing something totally unnecessary. (But it was a fun blog post! Kudos for that)
In which case, this is actually a much easier problem to solve since all you need to do is detect any motion (instead of all this convoluted--although hilarious--deep learning task).
This is like the story where some government tried to come up with a super high-tech pen that works in the space (because regular pens don't work in zero gravity environment) when the final answer was just a pencil.
So the lesson here is: there are many ways to solve a problem, and if you don't know what problem you're trying to solve and you don't know your customer well enough, you will end up wasting a lot of resource doing something totally unnecessary. (But it was a fun blog post! Kudos for that)