Earlier this year I was working on a personal project where I was automating the adding of surgical masks to faces in supplied images (repo here) and an issue I was having was that depending on the direction someone was facing in an image, we’d need to apply the mask image facing the same direction otherwise the effect didn’t look convincing at all.
To help me understand how to apply DLIB’s methodology for head pose detection where I found the documentation not simple, I created this notebook which demonstrates the steps and gets it working in Jupityr.
