implemented memory profiler

README.md

@@ -238,16 +238,23 @@ Below is output of `human.defaults` object
 Any property can be overriden by passing user object during `human.detect()`  
 Note that user object and default configuration are merged using deep-merge, so you do not need to redefine entire configuration  
 
+All configuration details can be changed in real-time!  
+
 Configurtion object is large, but typically you only need to modify few values:
 
 - `enabled`: Choose which models to use
 - `modelPath`: Update as needed to reflect your application's relative path
 
-
 ```js
 config = {
   backend: 'webgl',          // select tfjs backend to use
   console: true,             // enable debugging output to console
+  profile: true,             // enable tfjs profiling
+                             // this has significant performance impact, only enable for debugging purposes
+                             // currently only implemented for age,gender,emotion models
+  deallocate: true,          // aggresively deallocate gpu memory after each usage
+                             // only valid for webgl backend and only during first call, cannot be changed unless library is reloaded
+                             // this has significant performance impact, only enable on low-memory devices
   scoped: false,             // enable scoped runs
                              // some models *may* have memory leaks, this wrapps everything in a local scope at a cost of performance
                              // typically not needed
@@ -415,6 +422,35 @@ result = {
 
 <hr>
 
+## Profile
+
+If `config.profile` is enabled, call to `human.profile()` will return detailed profiling data from the last detect invokation.
+
+example:
+```js
+  result = {
+    {age: {…}, gender: {…}, emotion: {…}}
+      age:
+        timeKernelOps: 53.78892800000002
+        newBytes: 4
+        newTensors: 1
+        numKernelOps: 341
+        peakBytes: 46033948
+        largestKernelOps: Array(5)
+          0: {name: "Reshape", bytesAdded: 107648, totalBytesSnapshot: 46033948, tensorsAdded: 1, totalTensorsSnapshot: 1149, …}
+          1: {name: "Reshape", bytesAdded: 0, totalBytesSnapshot: 45818652, tensorsAdded: 1, totalTensorsSnapshot: 1147, …}
+          2: {name: "Reshape", bytesAdded: 0, totalBytesSnapshot: 45633996, tensorsAdded: 1, totalTensorsSnapshot: 1148, …}
+          3: {name: "Reshape", bytesAdded: 0, totalBytesSnapshot: 45389376, tensorsAdded: 1, totalTensorsSnapshot: 1154, …}
+          4: {name: "Reshape", bytesAdded: 53824, totalBytesSnapshot: 45381776, tensorsAdded: 1, totalTensorsSnapshot: 1155, …}
+        slowestKernelOps: Array(5)
+          0: {name: "_FusedMatMul", bytesAdded: 12, totalBytesSnapshot: 44802280, tensorsAdded: 1, totalTensorsSnapshot: 1156, …}
+          1: {name: "_FusedMatMul", bytesAdded: 4, totalBytesSnapshot: 44727564, tensorsAdded: 1, totalTensorsSnapshot: 1152, …}
+          2: {name: "_FusedMatMul", bytesAdded: 12, totalBytesSnapshot: 44789100, tensorsAdded: 1, totalTensorsSnapshot: 1157, …}
+          3: {name: "Add", bytesAdded: 4, totalBytesSnapshot: 44788748, tensorsAdded: 1, totalTensorsSnapshot: 1158, …}
+          4: {name: "Add", bytesAdded: 4, totalBytesSnapshot: 44788748, tensorsAdded: 1, totalTensorsSnapshot: 1158, …}
+  }
+```
+
 ## Build
 
 If you want to modify the library and perform a full rebuild:  

config.js

@@ -4,6 +4,12 @@
 export default {
   backend: 'webgl',          // select tfjs backend to use
   console: true,             // enable debugging output to console
+  profile: true,             // enable tfjs profiling
+                             // this has significant performance impact, only enable for debugging purposes
+                             // currently only implemented for age,gender,emotion models
+  deallocate: true,          // aggresively deallocate gpu memory after each usage
+                             // only valid for webgl backend and only during first call, cannot be changed unless library is reloaded
+                             // this has significant performance impact, only enable on low-memory devices
   scoped: false,             // enable scoped runs
                              // some models *may* have memory leaks, this wrapps everything in a local scope at a cost of performance
                              // typically not needed

demo/browser.js

@@ -30,6 +30,8 @@ const ui = {
 // configuration overrides
 const config = {
   backend: 'webgl',
+  profile: false,
+  deallocate: false,
   wasm: { path: '../assets' },
   filter: {
     enabled: true,
@@ -199,6 +201,7 @@ function runHumanDetect(input, canvas) {
       human.detect(input, config).then((result) => {
         if (result.error) log(result.error);
         else drawResults(input, result, canvas);
+        if (config.profile) log('Profile data:', human.profile());
       });
     }
   }
@@ -269,6 +272,8 @@ function setupMenu() {
 
   menu.addHTML('<hr style="min-width: 200px; border-style: inset; border-color: dimgray">');
   menu.addList('Backend', ['cpu', 'webgl', 'wasm', 'webgpu'], config.backend, (val) => config.backend = val);
+  menu.addBool('Enable Profiler', config, 'profile');
+  menu.addBool('Memory Deallocator', config, 'deallocate');
   menu.addBool('Use Web Worker', ui, 'useWorker');
   menu.addHTML('<hr style="min-width: 200px; border-style: inset; border-color: dimgray">');
   menu.addLabel('Enabled Models');

demo/menu.js

@@ -12,6 +12,7 @@ let theme = {
   checkboxOff: 'lightcoral',
   rangeBackground: 'lightblue',
   rangeLabel: 'white',
+  chartColor: 'lightblue',
 };
 
 function createCSS() {
@@ -69,8 +70,6 @@ class Menu {
     instance++;
     this._maxFPS = 0;
     this.hidden = 0;
-    this.chartFGcolor = 'lightblue';
-    this.chartBGcolor = 'lightgray';
   }
 
   createMenu(parent, title = '', position = { top: null, left: null, bottom: null, right: null }) {
@@ -256,13 +255,12 @@ class Menu {
     else this.addValue(title, val);
   }
 
-  addChart(title, id, width = 200, height = 40, fgColor, bgColor) {
+  addChart(title, id, width = 200, height = 40, color) {
-    if (fgColor) this.chartFGcolor = fgColor;
+    if (color) theme.chartColor = color;
-    if (bgColor) this.chartBGcolor = bgColor;
     const el = document.createElement('div');
     el.className = 'menu-item menu-chart-title';
     el.id = this.newID;
-    el.innerHTML = `<font color=${this.chartFGcolor}>${title}</font><canvas id="menu-canvas-${id}" class="menu-chart-canvas" width="${width}px" height="${height}px"></canvas>`;
+    el.innerHTML = `<font color=${theme.chartColor}>${title}</font><canvas id="menu-canvas-${id}" class="menu-chart-canvas" width="${width}px" height="${height}px"></canvas>`;
     this.container.appendChild(el);
   }
 
@@ -272,18 +270,18 @@ class Menu {
     const canvas = document.getElementById(`menu-canvas-${id}`);
     if (!canvas) return;
     const ctx = canvas.getContext('2d');
-    ctx.fillStyle = this.chartBGcolor;
+    ctx.fillStyle = theme.background;
     ctx.fillRect(0, 0, canvas.width, canvas.height);
     const width = canvas.width / values.length;
     const max = 1 + Math.max(...values);
     const height = canvas.height / max;
     for (const i in values) {
       const gradient = ctx.createLinearGradient(0, (max - values[i]) * height, 0, 0);
-      gradient.addColorStop(0.1, this.chartFGcolor);
+      gradient.addColorStop(0.1, theme.chartColor);
-      gradient.addColorStop(0.4, this.chartBGcolor);
+      gradient.addColorStop(0.4, theme.background);
       ctx.fillStyle = gradient;
       ctx.fillRect(i * width, 0, width - 4, canvas.height);
-      ctx.fillStyle = this.chartBGcolor;
+      ctx.fillStyle = theme.background;
       ctx.font = `${width / 1.4}px "Segoe UI"`;
       ctx.fillText(Math.round(values[i]), i * width + 1, canvas.height - 1, width - 1);
     }

dist/human.esm-nobundle.js

@@ -3859,9 +3859,37 @@ var require_facemesh = __commonJS((exports) => {
   exports.triangulation = triangulation;
 });
 
+// src/profile.js
+var require_profile = __commonJS((exports) => {
+  const profileData = {};
+  function profile2(name, data) {
+    if (!data || !data.kernels)
+      return;
+    const maxResults = 5;
+    const time = data.kernels.filter((a) => a.kernelTimeMs > 0).reduce((a, b) => a += b.kernelTimeMs, 0);
+    const slowest = data.kernels.map((a, i) => {
+      a.id = i;
+      return a;
+    }).filter((a) => a.kernelTimeMs > 0).sort((a, b) => b.kernelTimeMs - a.kernelTimeMs);
+    const largest = data.kernels.map((a, i) => {
+      a.id = i;
+      return a;
+    }).filter((a) => a.totalBytesSnapshot > 0).sort((a, b) => b.totalBytesSnapshot - a.totalBytesSnapshot);
+    if (slowest.length > maxResults)
+      slowest.length = maxResults;
+    if (largest.length > maxResults)
+      largest.length = maxResults;
+    const res = {newBytes: data.newBytes, newTensors: data.newTensors, peakBytes: data.peakBytes, numKernelOps: data.kernels.length, timeKernelOps: time, slowestKernelOps: slowest, largestKernelOps: largest};
+    profileData[name] = res;
+  }
+  exports.run = profile2;
+  exports.data = profileData;
+});
+
 // src/ssrnet/ssrnet.js
 var require_ssrnet = __commonJS((exports) => {
   const tf2 = require("@tensorflow/tfjs");
+  const profile2 = require_profile();
   const models = {};
   let last = {age: 0, gender: ""};
   let frame = 0;
@@ -3887,12 +3915,23 @@ var require_ssrnet = __commonJS((exports) => {
     const promises = [];
     let ageT;
     let genderT;
+    const obj = {};
+    if (!config.profile) {
       if (config.face.age.enabled)
         promises.push(ageT = models.age.predict(enhance));
       if (config.face.gender.enabled)
         promises.push(genderT = models.gender.predict(enhance));
       await Promise.all(promises);
-    const obj = {};
+    } else {
+      const profileAge = config.face.age.enabled ? await tf2.profile(() => models.age.predict(enhance)) : {};
+      ageT = profileAge.result.clone();
+      profileAge.result.dispose();
+      profile2.run("age", profileAge);
+      const profileGender = config.face.gender.enabled ? await tf2.profile(() => models.gender.predict(enhance)) : {};
+      genderT = profileGender.result.clone();
+      profileGender.result.dispose();
+      profile2.run("gender", profileGender);
+    }
     if (ageT) {
       const data = await ageT.data();
       obj.age = Math.trunc(10 * data[0]) / 10;
@@ -3919,6 +3958,7 @@ var require_ssrnet = __commonJS((exports) => {
 // src/emotion/emotion.js
 var require_emotion = __commonJS((exports) => {
   const tf2 = require("@tensorflow/tfjs");
+  const profile2 = require_profile();
   const annotations = ["angry", "discust", "fear", "happy", "sad", "surpise", "neutral"];
   const models = {};
   let last = [];
@@ -3950,14 +3990,22 @@ var require_emotion = __commonJS((exports) => {
     blueNorm.dispose();
     const obj = [];
     if (config.face.emotion.enabled) {
+      let data;
+      if (!config.profile) {
         const emotionT = await models.emotion.predict(grayscale);
-      const data = await emotionT.data();
+        data = await emotionT.data();
+        tf2.dispose(emotionT);
+      } else {
+        const profileData = await tf2.profile(() => models.emotion.predict(grayscale));
+        data = await profileData.result.data();
+        profileData.result.dispose();
+        profile2.run("emotion", profileData);
+      }
       for (let i = 0; i < data.length; i++) {
         if (multiplier * data[i] > config.face.emotion.minConfidence)
           obj.push({score: Math.min(0.99, Math.trunc(100 * multiplier * data[i]) / 100), emotion: annotations[i]});
       }
       obj.sort((a, b) => b.score - a.score);
-      tf2.dispose(emotionT);
     }
     tf2.dispose(grayscale);
     last = obj;
@@ -3974,8 +4022,6 @@ var require_modelBase = __commonJS((exports) => {
     constructor(model, outputStride) {
       this.model = model;
       this.outputStride = outputStride;
-      const inputShape = this.model.inputs[0].shape;
-      tf2.util.assert(inputShape[1] === -1 && inputShape[2] === -1, () => `Input shape [${inputShape[1]}, ${inputShape[2]}] must both be equal to or -1`);
     }
     predict(input) {
       return tf2.tidy(() => {
@@ -5682,6 +5728,8 @@ var require_config = __commonJS((exports) => {
   var config_default = {
     backend: "webgl",
     console: true,
+    profile: true,
+    deallocate: true,
     scoped: false,
     videoOptimized: true,
     filter: {
@@ -5777,7 +5825,7 @@ var require_config = __commonJS((exports) => {
 var require_package = __commonJS((exports, module) => {
   module.exports = {
     name: "@vladmandic/human",
-    version: "0.5.2",
+    version: "0.5.3",
     description: "human: 3D Face Detection, Iris Tracking and Age & Gender Prediction",
     sideEffects: false,
     main: "dist/human.node.js",
@@ -5849,6 +5897,7 @@ const emotion = require_emotion();
 const posenet = require_posenet();
 const handpose = require_handpose();
 const fxImage = require_imagefx();
+const profile = require_profile();
 const defaults = require_config().default;
 const app = require_package();
 let first = true;
@@ -5920,6 +5969,11 @@ class Human {
     if (msg && this.config.console)
       console.log("Human:", ...msg);
   }
+  profile() {
+    if (this.config.profile)
+      return profile.data;
+    return {};
+  }
   analyze(...msg) {
     if (!this.analyzeMemoryLeaks)
       return;
@@ -5961,13 +6015,14 @@ class Human {
   async checkBackend() {
     if (tf.getBackend() !== this.config.backend) {
       this.state = "backend";
-      if (this.config.backend in tf.engine().registry) {
       this.log("Setting backend:", this.config.backend);
       await tf.setBackend(this.config.backend);
-        await tf.ready();
+      tf.enableProdMode();
-      } else {
+      if (this.config.deallocate && this.config.backend === "webgl") {
-        this.log("Backend not registred:", this.config.backend);
+        this.log("Changing WebGL: WEBGL_DELETE_TEXTURE_THRESHOLD:", this.config.deallocate);
+        tf.ENV.set("WEBGL_DELETE_TEXTURE_THRESHOLD", this.config.deallocate ? 0 : -1);
       }
+      await tf.ready();
     }
   }
   tfImage(input) {

dist/human.esm-nobundle.json

@@ -1,7 +1,7 @@
 {
   "inputs": {
     "config.js": {
-      "bytes": 6295,
+      "bytes": 6870,
       "imports": []
     },
     "package.json": {
@@ -9,8 +9,12 @@
       "imports": []
     },
     "src/emotion/emotion.js": {
-      "bytes": 1646,
+      "bytes": 1958,
-      "imports": []
+      "imports": [
+        {
+          "path": "src/profile.js"
+        }
+      ]
     },
     "src/facemesh/blazeface.js": {
       "bytes": 7161,
@@ -116,7 +120,7 @@
       "imports": []
     },
     "src/human.js": {
-      "bytes": 13072,
+      "bytes": 13664,
       "imports": [
         {
           "path": "src/facemesh/facemesh.js"
@@ -136,6 +140,9 @@
         {
           "path": "src/imagefx.js"
         },
+        {
+          "path": "src/profile.js"
+        },
         {
           "path": "config.js"
         },
@@ -190,7 +197,7 @@
       "imports": []
     },
     "src/posenet/modelBase.js": {
-      "bytes": 1719,
+      "bytes": 1512,
       "imports": []
     },
     "src/posenet/modelMobileNet.js": {
@@ -251,16 +258,24 @@
         }
       ]
     },
-    "src/ssrnet/ssrnet.js": {
+    "src/profile.js": {
-      "bytes": 1574,
+      "bytes": 1004,
       "imports": []
+    },
+    "src/ssrnet/ssrnet.js": {
+      "bytes": 2115,
+      "imports": [
+        {
+          "path": "src/profile.js"
+        }
+      ]
     }
   },
   "outputs": {
     "dist/human.esm-nobundle.js.map": {
       "imports": [],
       "inputs": {},
-      "bytes": 250207
+      "bytes": 254266
     },
     "dist/human.esm-nobundle.js": {
       "imports": [],
@@ -289,14 +304,17 @@
         "src/facemesh/facemesh.js": {
           "bytesInOutput": 2661
         },
+        "src/profile.js": {
+          "bytesInOutput": 1092
+        },
         "src/ssrnet/ssrnet.js": {
-          "bytesInOutput": 1744
+          "bytesInOutput": 2310
         },
         "src/emotion/emotion.js": {
-          "bytesInOutput": 1718
+          "bytesInOutput": 2044
         },
         "src/posenet/modelBase.js": {
-          "bytesInOutput": 1118
+          "bytesInOutput": 910
         },
         "src/posenet/modelMobileNet.js": {
           "bytesInOutput": 504
@@ -350,19 +368,19 @@
           "bytesInOutput": 20195
         },
         "config.js": {
-          "bytesInOutput": 2230
+          "bytesInOutput": 2271
         },
         "package.json": {
           "bytesInOutput": 3012
         },
         "src/human.js": {
-          "bytesInOutput": 11537
+          "bytesInOutput": 11796
         },
         "src/human.js": {
           "bytesInOutput": 0
         }
       },
-      "bytes": 156000
+      "bytes": 158095
     }
   }
 }

dist/human.esm.json

@@ -1,11 +1,11 @@
 {
   "inputs": {
     "config.js": {
-      "bytes": 6295,
+      "bytes": 6870,
       "imports": []
     },
     "node_modules/@tensorflow/tfjs-backend-cpu/dist/tf-backend-cpu.node.js": {
-      "bytes": 297728,
+      "bytes": 272720,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js"
@@ -74,7 +74,7 @@
       ]
     },
     "node_modules/@tensorflow/tfjs-backend-webgl/dist/tf-backend-webgl.node.js": {
-      "bytes": 607535,
+      "bytes": 571410,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js"
@@ -82,7 +82,7 @@
       ]
     },
     "node_modules/@tensorflow/tfjs-converter/dist/tf-converter.node.js": {
-      "bytes": 312103,
+      "bytes": 294510,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js"
@@ -90,7 +90,7 @@
       ]
     },
     "node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js": {
-      "bytes": 1313261,
+      "bytes": 1305668,
       "imports": [
         {
           "path": "empty:/home/vlado/dev/human/node_modules/node-fetch/browser.js"
@@ -104,7 +104,7 @@
       ]
     },
     "node_modules/@tensorflow/tfjs-data/dist/tf-data.node.js": {
-      "bytes": 217014,
+      "bytes": 217016,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js"
@@ -153,10 +153,13 @@
       "imports": []
     },
     "src/emotion/emotion.js": {
-      "bytes": 1646,
+      "bytes": 1958,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs/dist/tf.node.js"
+        },
+        {
+          "path": "src/profile.js"
         }
       ]
     },
@@ -291,7 +294,7 @@
       "imports": []
     },
     "src/human.js": {
-      "bytes": 13072,
+      "bytes": 13664,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs/dist/tf.node.js"
@@ -314,6 +317,9 @@
         {
           "path": "src/imagefx.js"
         },
+        {
+          "path": "src/profile.js"
+        },
         {
           "path": "config.js"
         },
@@ -368,7 +374,7 @@
       "imports": []
     },
     "src/posenet/modelBase.js": {
-      "bytes": 1719,
+      "bytes": 1512,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs/dist/tf.node.js"
@@ -439,11 +445,18 @@
         }
       ]
     },
+    "src/profile.js": {
+      "bytes": 1004,
+      "imports": []
+    },
     "src/ssrnet/ssrnet.js": {
-      "bytes": 1574,
+      "bytes": 2115,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs/dist/tf.node.js"
+        },
+        {
+          "path": "src/profile.js"
         }
       ]
     },
@@ -468,7 +481,7 @@
     "dist/human.esm.js.map": {
       "imports": [],
       "inputs": {},
-      "bytes": 5123513
+      "bytes": 4999432
     },
     "dist/human.esm.js": {
       "imports": [],
@@ -483,19 +496,19 @@
           "bytesInOutput": 44
         },
         "node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js": {
-          "bytesInOutput": 1010337
+          "bytesInOutput": 1004766
         },
         "node_modules/@tensorflow/tfjs-layers/dist/tf-layers.node.js": {
           "bytesInOutput": 514491
         },
         "node_modules/@tensorflow/tfjs-converter/dist/tf-converter.node.js": {
-          "bytesInOutput": 258960
+          "bytesInOutput": 246275
         },
         "empty:/home/vlado/dev/human/node_modules/string_decoder/lib/string_decoder.js": {
           "bytesInOutput": 52
         },
         "node_modules/@tensorflow/tfjs-data/dist/tf-data.node.js": {
-          "bytesInOutput": 129585
+          "bytesInOutput": 129586
         },
         "node_modules/@tensorflow/tfjs-backend-cpu/node_modules/seedrandom/lib/alea.js": {
           "bytesInOutput": 2112
@@ -522,10 +535,10 @@
           "bytesInOutput": 458
         },
         "node_modules/@tensorflow/tfjs-backend-cpu/dist/tf-backend-cpu.node.js": {
-          "bytesInOutput": 272394
+          "bytesInOutput": 246000
         },
         "node_modules/@tensorflow/tfjs-backend-webgl/dist/tf-backend-webgl.node.js": {
-          "bytesInOutput": 561656
+          "bytesInOutput": 527935
         },
         "node_modules/@tensorflow/tfjs/dist/tf.node.js": {
           "bytesInOutput": 3025
@@ -554,14 +567,17 @@
         "src/facemesh/facemesh.js": {
           "bytesInOutput": 2651
         },
+        "src/profile.js": {
+          "bytesInOutput": 1092
+        },
         "src/ssrnet/ssrnet.js": {
-          "bytesInOutput": 1734
+          "bytesInOutput": 2300
         },
         "src/emotion/emotion.js": {
-          "bytesInOutput": 1708
+          "bytesInOutput": 2034
         },
         "src/posenet/modelBase.js": {
-          "bytesInOutput": 1108
+          "bytesInOutput": 900
         },
         "src/posenet/modelMobileNet.js": {
           "bytesInOutput": 494
@@ -615,19 +631,19 @@
           "bytesInOutput": 20195
         },
         "config.js": {
-          "bytesInOutput": 2230
+          "bytesInOutput": 2271
         },
         "package.json": {
           "bytesInOutput": 3012
         },
         "src/human.js": {
-          "bytesInOutput": 11527
+          "bytesInOutput": 11786
         },
         "src/human.js": {
           "bytesInOutput": 0
         }
       },
-      "bytes": 2924964
+      "bytes": 2848689
     }
   }
 }

dist/human.json

@@ -1,11 +1,11 @@
 {
   "inputs": {
     "config.js": {
-      "bytes": 6295,
+      "bytes": 6870,
       "imports": []
     },
     "node_modules/@tensorflow/tfjs-backend-cpu/dist/tf-backend-cpu.node.js": {
-      "bytes": 297728,
+      "bytes": 272720,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js"
@@ -74,7 +74,7 @@
       ]
     },
     "node_modules/@tensorflow/tfjs-backend-webgl/dist/tf-backend-webgl.node.js": {
-      "bytes": 607535,
+      "bytes": 571410,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js"
@@ -82,7 +82,7 @@
       ]
     },
     "node_modules/@tensorflow/tfjs-converter/dist/tf-converter.node.js": {
-      "bytes": 312103,
+      "bytes": 294510,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js"
@@ -90,7 +90,7 @@
       ]
     },
     "node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js": {
-      "bytes": 1313261,
+      "bytes": 1305668,
       "imports": [
         {
           "path": "empty:/home/vlado/dev/human/node_modules/node-fetch/browser.js"
@@ -104,7 +104,7 @@
       ]
     },
     "node_modules/@tensorflow/tfjs-data/dist/tf-data.node.js": {
-      "bytes": 217014,
+      "bytes": 217016,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js"
@@ -153,10 +153,13 @@
       "imports": []
     },
     "src/emotion/emotion.js": {
-      "bytes": 1646,
+      "bytes": 1958,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs/dist/tf.node.js"
+        },
+        {
+          "path": "src/profile.js"
         }
       ]
     },
@@ -291,7 +294,7 @@
       "imports": []
     },
     "src/human.js": {
-      "bytes": 13072,
+      "bytes": 13664,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs/dist/tf.node.js"
@@ -314,6 +317,9 @@
         {
           "path": "src/imagefx.js"
         },
+        {
+          "path": "src/profile.js"
+        },
         {
           "path": "config.js"
         },
@@ -368,7 +374,7 @@
       "imports": []
     },
     "src/posenet/modelBase.js": {
-      "bytes": 1719,
+      "bytes": 1512,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs/dist/tf.node.js"
@@ -439,11 +445,18 @@
         }
       ]
     },
+    "src/profile.js": {
+      "bytes": 1004,
+      "imports": []
+    },
     "src/ssrnet/ssrnet.js": {
-      "bytes": 1574,
+      "bytes": 2115,
       "imports": [
         {
           "path": "node_modules/@tensorflow/tfjs/dist/tf.node.js"
+        },
+        {
+          "path": "src/profile.js"
         }
       ]
     },
@@ -468,7 +481,7 @@
     "dist/human.js.map": {
       "imports": [],
       "inputs": {},
-      "bytes": 5127370
+      "bytes": 5003247
     },
     "dist/human.js": {
       "imports": [],
@@ -483,19 +496,19 @@
           "bytesInOutput": 48
         },
         "node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js": {
-          "bytesInOutput": 1061246
+          "bytesInOutput": 1055423
         },
         "node_modules/@tensorflow/tfjs-layers/dist/tf-layers.node.js": {
           "bytesInOutput": 541660
         },
         "node_modules/@tensorflow/tfjs-converter/dist/tf-converter.node.js": {
-          "bytesInOutput": 273096
+          "bytesInOutput": 259681
         },
         "empty:/home/vlado/dev/human/node_modules/string_decoder/lib/string_decoder.js": {
           "bytesInOutput": 56
         },
         "node_modules/@tensorflow/tfjs-data/dist/tf-data.node.js": {
-          "bytesInOutput": 136972
+          "bytesInOutput": 136974
         },
         "node_modules/@tensorflow/tfjs-backend-cpu/node_modules/seedrandom/lib/alea.js": {
           "bytesInOutput": 2276
@@ -522,10 +535,10 @@
           "bytesInOutput": 490
         },
         "node_modules/@tensorflow/tfjs-backend-cpu/dist/tf-backend-cpu.node.js": {
-          "bytesInOutput": 284249
+          "bytesInOutput": 256749
         },
         "node_modules/@tensorflow/tfjs-backend-webgl/dist/tf-backend-webgl.node.js": {
-          "bytesInOutput": 581862
+          "bytesInOutput": 546737
         },
         "node_modules/@tensorflow/tfjs/dist/tf.node.js": {
           "bytesInOutput": 3189
@@ -554,14 +567,17 @@
         "src/facemesh/facemesh.js": {
           "bytesInOutput": 2775
         },
+        "src/profile.js": {
+          "bytesInOutput": 1140
+        },
         "src/ssrnet/ssrnet.js": {
-          "bytesInOutput": 1835
+          "bytesInOutput": 2420
         },
         "src/emotion/emotion.js": {
-          "bytesInOutput": 1796
+          "bytesInOutput": 2137
         },
         "src/posenet/modelBase.js": {
-          "bytesInOutput": 1165
+          "bytesInOutput": 954
         },
         "src/posenet/modelMobileNet.js": {
           "bytesInOutput": 519
@@ -615,16 +631,16 @@
           "bytesInOutput": 21613
         },
         "config.js": {
-          "bytesInOutput": 2424
+          "bytesInOutput": 2469
         },
         "package.json": {
           "bytesInOutput": 3144
         },
         "src/human.js": {
-          "bytesInOutput": 12832
+          "bytesInOutput": 13119
         }
       },
-      "bytes": 3070917
+      "bytes": 2991264
     }
   }
 }

dist/human.node-nobundle.js

@@ -3859,9 +3859,37 @@ var require_facemesh = __commonJS((exports2) => {
   exports2.triangulation = triangulation;
 });
 
+// src/profile.js
+var require_profile = __commonJS((exports2) => {
+  const profileData = {};
+  function profile2(name, data) {
+    if (!data || !data.kernels)
+      return;
+    const maxResults = 5;
+    const time = data.kernels.filter((a) => a.kernelTimeMs > 0).reduce((a, b) => a += b.kernelTimeMs, 0);
+    const slowest = data.kernels.map((a, i) => {
+      a.id = i;
+      return a;
+    }).filter((a) => a.kernelTimeMs > 0).sort((a, b) => b.kernelTimeMs - a.kernelTimeMs);
+    const largest = data.kernels.map((a, i) => {
+      a.id = i;
+      return a;
+    }).filter((a) => a.totalBytesSnapshot > 0).sort((a, b) => b.totalBytesSnapshot - a.totalBytesSnapshot);
+    if (slowest.length > maxResults)
+      slowest.length = maxResults;
+    if (largest.length > maxResults)
+      largest.length = maxResults;
+    const res = {newBytes: data.newBytes, newTensors: data.newTensors, peakBytes: data.peakBytes, numKernelOps: data.kernels.length, timeKernelOps: time, slowestKernelOps: slowest, largestKernelOps: largest};
+    profileData[name] = res;
+  }
+  exports2.run = profile2;
+  exports2.data = profileData;
+});
+
 // src/ssrnet/ssrnet.js
 var require_ssrnet = __commonJS((exports2) => {
   const tf2 = require("@tensorflow/tfjs");
+  const profile2 = require_profile();
   const models = {};
   let last = {age: 0, gender: ""};
   let frame = 0;
@@ -3887,12 +3915,23 @@ var require_ssrnet = __commonJS((exports2) => {
     const promises = [];
     let ageT;
     let genderT;
+    const obj = {};
+    if (!config.profile) {
       if (config.face.age.enabled)
         promises.push(ageT = models.age.predict(enhance));
       if (config.face.gender.enabled)
         promises.push(genderT = models.gender.predict(enhance));
       await Promise.all(promises);
-    const obj = {};
+    } else {
+      const profileAge = config.face.age.enabled ? await tf2.profile(() => models.age.predict(enhance)) : {};
+      ageT = profileAge.result.clone();
+      profileAge.result.dispose();
+      profile2.run("age", profileAge);
+      const profileGender = config.face.gender.enabled ? await tf2.profile(() => models.gender.predict(enhance)) : {};
+      genderT = profileGender.result.clone();
+      profileGender.result.dispose();
+      profile2.run("gender", profileGender);
+    }
     if (ageT) {
       const data = await ageT.data();
       obj.age = Math.trunc(10 * data[0]) / 10;
@@ -3919,6 +3958,7 @@ var require_ssrnet = __commonJS((exports2) => {
 // src/emotion/emotion.js
 var require_emotion = __commonJS((exports2) => {
   const tf2 = require("@tensorflow/tfjs");
+  const profile2 = require_profile();
   const annotations = ["angry", "discust", "fear", "happy", "sad", "surpise", "neutral"];
   const models = {};
   let last = [];
@@ -3950,14 +3990,22 @@ var require_emotion = __commonJS((exports2) => {
     blueNorm.dispose();
     const obj = [];
     if (config.face.emotion.enabled) {
+      let data;
+      if (!config.profile) {
         const emotionT = await models.emotion.predict(grayscale);
-      const data = await emotionT.data();
+        data = await emotionT.data();
+        tf2.dispose(emotionT);
+      } else {
+        const profileData = await tf2.profile(() => models.emotion.predict(grayscale));
+        data = await profileData.result.data();
+        profileData.result.dispose();
+        profile2.run("emotion", profileData);
+      }
       for (let i = 0; i < data.length; i++) {
         if (multiplier * data[i] > config.face.emotion.minConfidence)
           obj.push({score: Math.min(0.99, Math.trunc(100 * multiplier * data[i]) / 100), emotion: annotations[i]});
       }
       obj.sort((a, b) => b.score - a.score);
-      tf2.dispose(emotionT);
     }
     tf2.dispose(grayscale);
     last = obj;
@@ -3974,8 +4022,6 @@ var require_modelBase = __commonJS((exports2) => {
     constructor(model, outputStride) {
       this.model = model;
       this.outputStride = outputStride;
-      const inputShape = this.model.inputs[0].shape;
-      tf2.util.assert(inputShape[1] === -1 && inputShape[2] === -1, () => `Input shape [${inputShape[1]}, ${inputShape[2]}] must both be equal to or -1`);
     }
     predict(input) {
       return tf2.tidy(() => {
@@ -5682,6 +5728,8 @@ var require_config = __commonJS((exports2) => {
   var config_default = {
     backend: "webgl",
     console: true,
+    profile: true,
+    deallocate: true,
     scoped: false,
     videoOptimized: true,
     filter: {
@@ -5777,7 +5825,7 @@ var require_config = __commonJS((exports2) => {
 var require_package = __commonJS((exports2, module2) => {
   module2.exports = {
     name: "@vladmandic/human",
-    version: "0.5.2",
+    version: "0.5.3",
     description: "human: 3D Face Detection, Iris Tracking and Age & Gender Prediction",
     sideEffects: false,
     main: "dist/human.node.js",
@@ -5852,6 +5900,7 @@ const emotion = require_emotion();
 const posenet = require_posenet();
 const handpose = require_handpose();
 const fxImage = require_imagefx();
+const profile = require_profile();
 const defaults = require_config().default;
 const app = require_package();
 let first = true;
@@ -5923,6 +5972,11 @@ class Human {
     if (msg && this.config.console)
       console.log("Human:", ...msg);
   }
+  profile() {
+    if (this.config.profile)
+      return profile.data;
+    return {};
+  }
   analyze(...msg) {
     if (!this.analyzeMemoryLeaks)
       return;
@@ -5964,13 +6018,14 @@ class Human {
   async checkBackend() {
     if (tf.getBackend() !== this.config.backend) {
       this.state = "backend";
-      if (this.config.backend in tf.engine().registry) {
       this.log("Setting backend:", this.config.backend);
       await tf.setBackend(this.config.backend);
-        await tf.ready();
+      tf.enableProdMode();
-      } else {
+      if (this.config.deallocate && this.config.backend === "webgl") {
-        this.log("Backend not registred:", this.config.backend);
+        this.log("Changing WebGL: WEBGL_DELETE_TEXTURE_THRESHOLD:", this.config.deallocate);
+        tf.ENV.set("WEBGL_DELETE_TEXTURE_THRESHOLD", this.config.deallocate ? 0 : -1);
       }
+      await tf.ready();
     }
   }
   tfImage(input) {

dist/human.node.json

@@ -1,7 +1,7 @@
 {
   "inputs": {
     "config.js": {
-      "bytes": 6295,
+      "bytes": 6870,
       "imports": []
     },
     "package.json": {
@@ -9,8 +9,12 @@
       "imports": []
     },
     "src/emotion/emotion.js": {
-      "bytes": 1646,
+      "bytes": 1958,
-      "imports": []
+      "imports": [
+        {
+          "path": "src/profile.js"
+        }
+      ]
     },
     "src/facemesh/blazeface.js": {
       "bytes": 7161,
@@ -116,7 +120,7 @@
       "imports": []
     },
     "src/human.js": {
-      "bytes": 13072,
+      "bytes": 13664,
       "imports": [
         {
           "path": "src/facemesh/facemesh.js"
@@ -136,6 +140,9 @@
         {
           "path": "src/imagefx.js"
         },
+        {
+          "path": "src/profile.js"
+        },
         {
           "path": "config.js"
         },
@@ -190,7 +197,7 @@
       "imports": []
     },
     "src/posenet/modelBase.js": {
-      "bytes": 1719,
+      "bytes": 1512,
       "imports": []
     },
     "src/posenet/modelMobileNet.js": {
@@ -251,16 +258,24 @@
         }
       ]
     },
-    "src/ssrnet/ssrnet.js": {
+    "src/profile.js": {
-      "bytes": 1574,
+      "bytes": 1004,
       "imports": []
+    },
+    "src/ssrnet/ssrnet.js": {
+      "bytes": 2115,
+      "imports": [
+        {
+          "path": "src/profile.js"
+        }
+      ]
     }
   },
   "outputs": {
     "dist/human.node-nobundle.js.map": {
       "imports": [],
       "inputs": {},
-      "bytes": 263646
+      "bytes": 268314
     },
     "dist/human.node-nobundle.js": {
       "imports": [],
@@ -289,14 +304,17 @@
         "src/facemesh/facemesh.js": {
           "bytesInOutput": 2666
         },
+        "src/profile.js": {
+          "bytesInOutput": 1095
+        },
         "src/ssrnet/ssrnet.js": {
-          "bytesInOutput": 1748
+          "bytesInOutput": 2314
         },
         "src/emotion/emotion.js": {
-          "bytesInOutput": 1721
+          "bytesInOutput": 2047
         },
         "src/posenet/modelBase.js": {
-          "bytesInOutput": 1120
+          "bytesInOutput": 912
         },
         "src/posenet/modelMobileNet.js": {
           "bytesInOutput": 506
@@ -350,7 +368,7 @@
           "bytesInOutput": 20197
         },
         "config.js": {
-          "bytesInOutput": 2232
+          "bytesInOutput": 2273
         },
         "package.json": {
           "bytesInOutput": 3015
@@ -359,10 +377,10 @@
           "bytesInOutput": 47
         },
         "src/human.js": {
-          "bytesInOutput": 11537
+          "bytesInOutput": 11796
         }
       },
-      "bytes": 156154
+      "bytes": 158252
     }
   }
 }

src/emotion/emotion.js

@@ -1,4 +1,5 @@
 const tf = require('@tensorflow/tfjs');
+const profile = require('../profile.js');
 
 const annotations = ['angry', 'discust', 'fear', 'happy', 'sad', 'surpise', 'neutral'];
 const models = {};
@@ -33,13 +34,21 @@ async function predict(image, config) {
   blueNorm.dispose();
   const obj = [];
   if (config.face.emotion.enabled) {
+    let data;
+    if (!config.profile) {
       const emotionT = await models.emotion.predict(grayscale);
-    const data = await emotionT.data();
+      data = await emotionT.data();
+      tf.dispose(emotionT);
+    } else {
+      const profileData = await tf.profile(() => models.emotion.predict(grayscale));
+      data = await profileData.result.data();
+      profileData.result.dispose();
+      profile.run('emotion', profileData);
+    }
     for (let i = 0; i < data.length; i++) {
       if (multiplier * data[i] > config.face.emotion.minConfidence) obj.push({ score: Math.min(0.99, Math.trunc(100 * multiplier * data[i]) / 100), emotion: annotations[i] });
     }
     obj.sort((a, b) => b.score - a.score);
-    tf.dispose(emotionT);
   }
   tf.dispose(grayscale);
   last = obj;

src/human.js

@@ -5,6 +5,7 @@ const emotion = require('./emotion/emotion.js');
 const posenet = require('./posenet/posenet.js');
 const handpose = require('./handpose/handpose.js');
 const fxImage = require('./imagefx.js');
+const profile = require('./profile.js');
 const defaults = require('../config.js').default;
 const app = require('../package.json');
 
@@ -88,6 +89,11 @@ class Human {
     if (msg && this.config.console) console.log('Human:', ...msg);
   }
 
+  profile() {
+    if (this.config.profile) return profile.data;
+    return {};
+  }
+
   // helper function: measure tensor leak
   analyze(...msg) {
     if (!this.analyzeMemoryLeaks) return;
@@ -129,16 +135,27 @@ class Human {
   async checkBackend() {
     if (tf.getBackend() !== this.config.backend) {
       this.state = 'backend';
+      /* force backend reload
       if (this.config.backend in tf.engine().registry) {
-        this.log('Setting backend:', this.config.backend);
+        const backendFactory = tf.findBackendFactory(this.config.backend);
-        // const backendFactory = tf.findBackendFactory(backendName);
+        tf.removeBackend(this.config.backend);
-        // tf.removeBackend(backendName);
+        tf.registerBackend(this.config.backend, backendFactory);
-        // tf.registerBackend(backendName, backendFactory);
-        await tf.setBackend(this.config.backend);
-        await tf.ready();
       } else {
         this.log('Backend not registred:', this.config.backend);
       }
+      */
+      this.log('Setting backend:', this.config.backend);
+      await tf.setBackend(this.config.backend);
+      tf.enableProdMode();
+      /* debug mode is really too mcuh
+      if (this.config.profile) tf.enableDebugMode();
+      else tf.enableProdMode();
+      */
+      if (this.config.deallocate && this.config.backend === 'webgl') {
+        this.log('Changing WebGL: WEBGL_DELETE_TEXTURE_THRESHOLD:', this.config.deallocate);
+        tf.ENV.set('WEBGL_DELETE_TEXTURE_THRESHOLD', this.config.deallocate ? 0 : -1);
+      }
+      await tf.ready();
     }
   }
 

src/posenet/modelBase.js

@@ -4,8 +4,6 @@ class BaseModel {
   constructor(model, outputStride) {
     this.model = model;
     this.outputStride = outputStride;
-    const inputShape = this.model.inputs[0].shape;
-    tf.util.assert((inputShape[1] === -1) && (inputShape[2] === -1), () => `Input shape [${inputShape[1]}, ${inputShape[2]}] must both be equal to or -1`);
   }
 
   /**

src/profile.js

@@ -0,0 +1,24 @@
+const profileData = {};
+
+function profile(name, data) {
+  if (!data || !data.kernels) return;
+  const maxResults = 5;
+  const time = data.kernels
+    .filter((a) => a.kernelTimeMs > 0)
+    .reduce((a, b) => a += b.kernelTimeMs, 0);
+  const slowest = data.kernels
+    .map((a, i) => { a.id = i; return a; })
+    .filter((a) => a.kernelTimeMs > 0)
+    .sort((a, b) => b.kernelTimeMs - a.kernelTimeMs);
+  const largest = data.kernels
+    .map((a, i) => { a.id = i; return a; })
+    .filter((a) => a.totalBytesSnapshot > 0)
+    .sort((a, b) => b.totalBytesSnapshot - a.totalBytesSnapshot);
+  if (slowest.length > maxResults) slowest.length = maxResults;
+  if (largest.length > maxResults) largest.length = maxResults;
+  const res = { newBytes: data.newBytes, newTensors: data.newTensors, peakBytes: data.peakBytes, numKernelOps: data.kernels.length, timeKernelOps: time, slowestKernelOps: slowest, largestKernelOps: largest };
+  profileData[name] = res;
+}
+
+exports.run = profile;
+exports.data = profileData;

src/ssrnet/ssrnet.js

@@ -1,4 +1,5 @@
 const tf = require('@tensorflow/tfjs');
+const profile = require('../profile.js');
 
 const models = {};
 let last = { age: 0, gender: '' };
@@ -27,11 +28,23 @@ async function predict(image, config) {
   const promises = [];
   let ageT;
   let genderT;
+  const obj = {};
+
+  if (!config.profile) {
     if (config.face.age.enabled) promises.push(ageT = models.age.predict(enhance));
     if (config.face.gender.enabled) promises.push(genderT = models.gender.predict(enhance));
     await Promise.all(promises);
+  } else {
+    const profileAge = config.face.age.enabled ? await tf.profile(() => models.age.predict(enhance)) : {};
+    ageT = profileAge.result.clone();
+    profileAge.result.dispose();
+    profile.run('age', profileAge);
+    const profileGender = config.face.gender.enabled ? await tf.profile(() => models.gender.predict(enhance)) : {};
+    genderT = profileGender.result.clone();
+    profileGender.result.dispose();
+    profile.run('gender', profileGender);
+  }
 
-  const obj = {};
   if (ageT) {
     const data = await ageT.data();
     obj.age = Math.trunc(10 * data[0]) / 10;