This ML program tracks the face and detects the exact 3D coordinates of facial landmarks like eyelashes, eyebrows, chin, nose tip and 478 such points. Program leverages transfer learning from Google's mediapipe model, trained on over 30K manually annotated dataset which ensures high accuracy and efficiency.
Sam-Varghese/FaceLandmarkTracker
Implementation of mediapipe transfer learning to track live feed facial landmarks through OpenCV.
Jupyter Notebook