📘 New concepts: load face-api.js models, run detection every frame,
map bounding box coordinates to mirrored canvas space, draw rectangle around face.
💡 Offline setup: download weights from face-api weights,
put them in a folder named models and change MODEL_PATH = './models' below.