optimized blazeface anchors

demo/demo-esm.js

@@ -404,9 +404,9 @@ async function main() {
   await setupUI();
   // setup webcam
   await setupCamera();
+
   // or setup image
   // const input = await setupImage();
-  // setup output canvas from input object
 
   const msg = `Human ready: version: ${human.version} TensorFlow/JS version: ${human.tf.version_core}`;
   document.getElementById('log').innerText = msg;

dist/human-nobundle.cjs

@@ -18,18 +18,15 @@ var __export = (target, all) => {
 // src/facemesh/blazeface.js
 var require_blazeface = __commonJS((exports2) => {
   const tf2 = require("@tensorflow/tfjs");
-  const ANCHORS_CONFIG = {
-    strides: [8, 16],
-    anchors: [2, 6]
-  };
   const NUM_LANDMARKS = 6;
-  function generateAnchors(anchorSize, outputSpec) {
+  function generateAnchors(inputSize) {
+    const spec = {strides: [inputSize / 16, inputSize / 8], anchors: [2, 6]};
     const anchors = [];
-    for (let i = 0; i < outputSpec.strides.length; i++) {
-      const stride = outputSpec.strides[i];
-      const gridRows = Math.floor((anchorSize + stride - 1) / stride);
-      const gridCols = Math.floor((anchorSize + stride - 1) / stride);
-      const anchorsNum = outputSpec.anchors[i];
+    for (let i = 0; i < spec.strides.length; i++) {
+      const stride = spec.strides[i];
+      const gridRows = Math.floor((inputSize + stride - 1) / stride);
+      const gridCols = Math.floor((inputSize + stride - 1) / stride);
+      const anchorsNum = spec.anchors[i];
       for (let gridY = 0; gridY < gridRows; gridY++) {
         const anchorY = stride * (gridY + 0.5);
         for (let gridX = 0; gridX < gridCols; gridX++) {
@@ -83,9 +80,8 @@ var require_blazeface = __commonJS((exports2) => {
       this.blazeFaceModel = model;
       this.width = config2.detector.inputSize;
       this.height = config2.detector.inputSize;
-      this.anchorSize = config2.detector.anchorSize;
       this.maxFaces = config2.detector.maxFaces;
-      this.anchorsData = generateAnchors(config2.detector.anchorSize, ANCHORS_CONFIG);
+      this.anchorsData = generateAnchors(config2.detector.inputSize);
       this.anchors = tf2.tensor2d(this.anchorsData);
       this.inputSize = tf2.tensor1d([this.width, this.height]);
       this.iouThreshold = config2.detector.iouThreshold;
@@ -5054,9 +5050,8 @@ var require_config = __commonJS((exports2) => {
     face: {
       enabled: true,
       detector: {
-        modelPath: "../models/blazeface/tfhub/model.json",
-        anchorSize: 128,
-        inputSize: 128,
+        modelPath: "../models/blazeface/back/model.json",
+        inputSize: 256,
         maxFaces: 10,
         skipFrames: 10,
         minConfidence: 0.5,

src/config.js

@@ -4,9 +4,8 @@ export default {
   face: {
     enabled: true, // refers to detector, but since all other face modules rely on detector, it should be a global
     detector: {
-      modelPath: '../models/blazeface/tfhub/model.json', // can be blazeface-front or blazeface-back
-      anchorSize: 128, // fixed regardless of model
-      inputSize: 128, // fixed value: 128 for front and tfhub and 256 for back
+      modelPath: '../models/blazeface/back/model.json', // can be blazeface-front or blazeface-back
+      inputSize: 256, // fixed value: 128 for front and tfhub and 256 for back
       maxFaces: 10, // maximum number of faces detected in the input, should be set to the minimum number for performance
       skipFrames: 10, // how many frames to go without running the bounding box detector
       minConfidence: 0.5, // threshold for discarding a prediction

src/facemesh/blazeface.js

@@ -1,18 +1,15 @@
 const tf = require('@tensorflow/tfjs');
 
-const ANCHORS_CONFIG = {
-  strides: [8, 16],
-  anchors: [2, 6],
-};
-
 const NUM_LANDMARKS = 6;
-function generateAnchors(anchorSize, outputSpec) {
+
+function generateAnchors(inputSize) {
+  const spec = { strides: [inputSize / 16, inputSize / 8], anchors: [2, 6] };
   const anchors = [];
-  for (let i = 0; i < outputSpec.strides.length; i++) {
-    const stride = outputSpec.strides[i];
-    const gridRows = Math.floor((anchorSize + stride - 1) / stride);
-    const gridCols = Math.floor((anchorSize + stride - 1) / stride);
-    const anchorsNum = outputSpec.anchors[i];
+  for (let i = 0; i < spec.strides.length; i++) {
+    const stride = spec.strides[i];
+    const gridRows = Math.floor((inputSize + stride - 1) / stride);
+    const gridCols = Math.floor((inputSize + stride - 1) / stride);
+    const anchorsNum = spec.anchors[i];
     for (let gridY = 0; gridY < gridRows; gridY++) {
       const anchorY = stride * (gridY + 0.5);
       for (let gridX = 0; gridX < gridCols; gridX++) {
@@ -72,9 +69,8 @@ class BlazeFaceModel {
     this.blazeFaceModel = model;
     this.width = config.detector.inputSize;
     this.height = config.detector.inputSize;
-    this.anchorSize = config.detector.anchorSize;
     this.maxFaces = config.detector.maxFaces;
-    this.anchorsData = generateAnchors(config.detector.anchorSize, ANCHORS_CONFIG);
+    this.anchorsData = generateAnchors(config.detector.inputSize);
     this.anchors = tf.tensor2d(this.anchorsData);
     this.inputSize = tf.tensor1d([this.width, this.height]);
     this.iouThreshold = config.detector.iouThreshold;