Hello folks. I built a pushup counter with OpenCV + Clojure. Here is a video it working (please excuse my poor pushup form).
First the counter takes a raw image and subtracts out the background.
Then it averages the Ycoordinate values of all the white pixels. This average value corresponds to the vertical center of my body. If I plot this Y value across time as I do a pushup, it forms a nice sinusoidal curve
I then compute the slope of that curve by simply subtracting the last point with the new point. When that slope flips sign, I know my body is in the pushup position and the counter adds one.
Here is a video of that underlying process in action:
The display and voice synthesis is relatively easy. There are many libraries and APIs to do GUI and voice synthesis. I used OpenCV’s drawing function for the GUI and I used Mac’s
say terminal command for the verbal audio count. If I were to do it again, I would get my computer to speak in my own voice using the LyreBird voice neural net API
Anyway, if you are interested in checking out the source code, you can find it here. The pushup counter was part of a larger project I put together a few years ago and it requires a lot of setup to get running. Despite that, my code may help you with some of your own computer vision ideas.