FaceMesh
Face landmark detection
RUN
The FaceMesh by MediaPipe model detects 468 key face landmarks in real time. The model can be configured to detect up to 20 faces.
The model outputs 468 normalized coordinates, each corresponding to some landmark, but the model only outputs 6 key-points
The following 6-keypoints are detected:
- Right Eye Inner
- Right Eye Outer
- Left Eye Inner
- Left Eye Outer
- Lip Right Corner
- Lip Left Corner
The data is exported as a CSV file structured as follows:
| Frame | right_eye_inner_x_pos | right_eye_inner_y_pos | right_eye_outer_x_pos | right_eye_inner_x_pos | left_eye_inner_x_pos | left_eye_inner_y_pos | left_eye_outer_x_pos | left_eye_inner_x_pos | lip_right_corner_x_pos | lip_right_corner_y_pos | lip_left_corner_x_pos | lup_left_corner_y_pos | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 11 | 251.2331 | 239.0843 | 208.1723 | 238.9472 | 290.5730 | 240.7219 | 332.6023 | 242.9274 | 230.8592 | 318.2940 | 305.4467 | 320.0349 |
| 1 | 32 | 251.1341 | 239.4022 | 208.9237 | 237.7281 | 290.0012 | 241.3881 | 332.2391 | 241.1928 | 231.3710 | 318.2001 | 305.1028 | 320.7389 |
The frame column represents the frame number in the video, the timestamp column represents the time in milliseconds that have transpired since the model started running (upon recording data), and the other columns give the coordinates (x, y) of the detected features.
- maxNumberOfFaces - the maximum number of faces for the model to detect
- refineLandMarks - whether or not to track the pupils
- minDetectionConfidence - Minimum confidence value ([0.0, 1.0]) from the face detection model for the detection to be considered successful. Default to 0.5.
- minTrackingConfidence - Minimum confidence value ([0.0, 1.0]) from the landmark-tracking model for the face landmarks to be considered tracked successfully, or otherwise face detection will be invoked automatically on the next input image. Setting it to a higher value can increase robustness of the solution, at the expense of a higher latency.