Face mesh detection concepts

Face mesh info contains two parts:

  • 468 3D points: Each point has a unique ID, ranging from 0 to 467, associated with a fixed position on the detected face. For each point:

    • x and y are pixel coordinates of the detected face.
    • z is the depth info scaled to image size, such that z-values are more negative when the face is closer to the camera. The origin is the average depth of all 468 points.
  • Triangle info: This is used to represent a logical triangle surface in the detected face. Each triangle contains three 3D points. For example, Points #0, #37, and #164 construct a small triangle area between the nose and lips.

Example of face mesh info, click for zoomable image