Hi HN. I built this computer vision pipeline recently to detect when I am wasting my life away on my phone. I've been wanting to dip my toes into computer vision and this ended up being a really fun project that works surprisingly well.
Uses YOLOv11 for pose estimation and object detection as well as OpenCV.
Were you not able to detect the doomscrolling when you were doing it before building this? Just curious why a person would need to be told when they're doing a behavior that makes them feel bad.