To track web cam movement, I used
Pose Net, a machine learning model that allows for Real-time Human Pose Estimation (shout out to
ml5). To control movement, I simply calculate directional deviations from your initial face placement, and move the sperm cell in the direction with the largest magnitude. Because everything is done in the browser, your web-cam data never leaves your computer, and is never stored anywhere.
All the processing stuff was handled using
p5.