Facial keypoints detection and tracking is an interesting field in computer vision. Deep learning models solves this problem with an amazing accuracy. However, in this repo, I have tried to implement the facial keypoints tracking using the concept of Optical flow in real-time.
There are several algorithms proposed to solve the problem of Optical flow. I have used the Lucas-kanade method to track the landmarks.
Steps:
- Input the number of points to track (Say, x).
- Select the x points on the live video (Try to remain still and then select).
- The selected points will be tracked on the live video.