let video;

let predictions = [];

​

function setup() {

  createCanvas(640, 480);

  video = createCapture(VIDEO);

  video.size(width, height);

​

  facemesh = ml5.facemesh(video, modelReady);

​

  // This sets up an event that fills the global variable "predictions"

  // with an array every time new predictions are made

  facemesh.on("predict", results => {

    predictions = results;

  });

​

  // Hide the video element, and just show the canvas

  video.hide();

}

​

function modelReady() {

  console.log("Model ready!");

}

​

function draw() {

  image(video, 0, 0, width, height);

​

  // We can call both functions to draw all keypoints

  drawKeypoints();

}

​

// A function to draw ellipses over the detected keypoints

function drawKeypoints() {

  for (let i = 0; i < predictions.length; i += 1) {

    const keypoints = predictions[i].scaledMesh;

​

    // Draw facial keypoints.

    for (let j = 0; j < keypoints.length; j += 1) {

      const [x, y, z] = keypoints[j];

      noStroke();

      const shade = ((z+10)*10 -millis()/2 %255 + 255) %255;

      fill(shade, shade, 255);

      ellipse(x, y, 5, 5);

    }

  }

}

​

function drawMesh() {

}

​