I learnt something from every single project I made, but this one... it was really different, new language, new library, new hardware.

there's an infopoint in a place, that was really hard to use (ball mouse over a monitor)

make the screen be touch.

Developing the solution:
- after a bit of research I find out there's a library/project called OpenCV
- there are example programs that detect hands written in C++ (which I know) and Python (which I don't know)
- the whole infopoint works on a raspberry PI, with raspbian (I'm not new to linux, but it's somehow different, plus tons of customization)

So I spend like 3 weeks trying to understand how to make it work, at first, the webcam was on the TV and you could move the mouse just by shaking your hand, but it didn't work too well, so we tried making the webcam look at the screen and then calculate the differences between "no-hands" and "user-hand", but should have been calibrated, wasn't too precise... dropped solution.

put the webcam 30cm above the screen, let it just analyse a few centimeters of sight from the screen and it worked flawlessly, BUT it could just recognise the horizontal axis => had to rework the infopoint UI to make it dumb-easy

It all finally worked, I learnt python, openCV, a bit of photography

Then hated it all and decided to never do that again

Add Comment