Set up testing

.babelrc

@@ -0,0 +1,3 @@
+{
+  "presets": ["@babel/preset-env"]
+}

CONTRIBUTING.md

@@ -71,6 +71,15 @@ yarn run build
 
 This will create a production version of the library in `/dist` directory.
 
+
+## Unit Tests
+
+To run the unit tests, run the following command
+
+```
+yarn test
+```
+
 ## Making Releases
 
 _This section is a temporary guide for contributors who wants to make a alpha release manually._

babel.config.json

@@ -0,0 +1,3 @@
+{
+  "presets": ["@babel/preset-env"]
+}

jest.config.js

@@ -0,0 +1,19 @@
+/**
+ * For a detailed explanation regarding each configuration property, visit:
+ * https://jestjs.io/docs/configuration
+ */
+
+/** @type {import('jest').Config} */
+const config = {
+  collectCoverage: true,
+  coverageDirectory: "coverage",
+  coverageProvider: "v8",
+  globalSetup: "./setupTests.js",
+  passWithNoTests: true,
+  testEnvironment: "jsdom",
+  testEnvironmentOptions: {
+    resources: "usable", // Load image resources
+  },
+};
+
+module.exports = config;

package.json

@@ -8,7 +8,8 @@
     "build": "webpack --config webpack.config.js --mode production",
     "start": "webpack serve --config webpack.config.js --mode development",
     "format": "prettier --write \"**/*.js\"",
-    "postinstall": "patch-package"
+    "postinstall": "patch-package",
+    "test": "jest"
   },
   "files": [
     "dist"
@@ -26,8 +27,17 @@
     "url": "https://github.com/ml5js/ml5-next-gen/issues"
   },
   "devDependencies": {
+    "@babel/core": "^7.23.2",
+    "@babel/preset-env": "^7.23.2",
+    "@tensorflow/tfjs-backend-webgpu": "^4.17.0",
+    "@tensorflow/tfjs-node": "^4.17.0",
     "all-contributors-cli": "^6.26.1",
+    "babel-jest": "^29.7.0",
+    "canvas": "^2.11.2",
+    "cross-fetch": "^4.0.0",
     "html-webpack-plugin": "^5.5.3",
+    "jest": "^29.7.0",
+    "jest-environment-jsdom": "^29.7.0",
     "patch-package": "^8.0.0",
     "postinstall-postinstall": "^2.1.0",
     "prettier": "2.8.8",
@@ -50,6 +60,7 @@
     "@tensorflow/tfjs": "^4.2.0",
     "@tensorflow/tfjs-vis": "^1.5.1",
     "axios": "^1.3.4",
+    "canvas": "^2.11.2",
     "webpack-merge": "^5.9.0"
-}
+  }
 }

setupTests.js

@@ -0,0 +1,27 @@
+const { ImageData } = require("canvas");
+import '@tensorflow/tfjs-node'; // loads the tensorflow/node backend to the registry
+import crossFetch from 'cross-fetch';
+import * as tf from '@tensorflow/tfjs';
+
+async function setupTests() {
+
+  console.log("Beginning setup");
+
+  await tf.setBackend('tensorflow');
+  tf.env().set('IS_BROWSER', false);
+  tf.env().set('IS_NODE', true);
+
+  // Use the node-canvas ImageData polyfill
+  if (!global.ImageData) {
+    global.ImageData = ImageData;
+  }
+
+  // Use cross-fetch as a polyfill for the browser fetch
+  if (!global.fetch) {
+    global.fetch = crossFetch;
+  }
+
+  console.log("Setup complete");
+}
+
+module.exports = setupTests;

src/BodyPose/index.js

@@ -215,7 +215,7 @@ class BodyPose {
     this.model = await poseDetection.createDetector(pipeline, modelConfig);
 
     // for compatibility with p5's preload()
-    if (this.p5PreLoadExists) window._decrementPreload();
+    if (this.p5PreLoadExists()) window._decrementPreload();
 
     return this;
   }

src/BodyPose/index.test.js

@@ -1,60 +1,63 @@
-// Copyright (c) 2018 ml5
+// Copyright (c) 2018-2024 ml5
 //
 // This software is released under the MIT License.
 // https://opensource.org/licenses/MIT
 
 import { asyncLoadImage } from "../utils/testingUtils";
-import poseNet from "./index";
+import bodyPose from "./index";
+import crossFetch from "cross-fetch";
 
 const POSENET_IMG =
   "https://github.com/ml5js/ml5-adjacent/raw/master/02_ImageClassification_Video/starter.png";
 
-const POSENET_DEFAULTS = {
-  architecture: "MobileNetV1",
-  outputStride: 16,
-  flipHorizontal: false,
-  minConfidence: 0.5,
-  maxPoseDetections: 5,
-  scoreThreshold: 0.5,
-  nmsRadius: 20,
-  detectionType: "multiple",
-  inputResolution: 256,
-  multiplier: 0.75,
-  quantBytes: 2,
-};
-
-describe("PoseNet", () => {
-  let net;
+describe("bodypose", () => {
+  let myBodyPose;
+  let image;
 
   beforeAll(async () => {
-    jest.setTimeout(10000);
-    net = await poseNet();
+
+    // TODO: this should not be necessary! Should already be handled by setupTests.js.
+    if (!global.fetch) {
+      global.fetch = crossFetch;
+    }
+
+    myBodyPose = bodyPose();
+    await myBodyPose.ready;
+
+    image = await asyncLoadImage(POSENET_IMG);
   });
 
-  it("instantiates poseNet", () => {
-    expect(net.architecture).toBe(POSENET_DEFAULTS.architecture);
-    expect(net.outputStride).toBe(POSENET_DEFAULTS.outputStride);
-    expect(net.inputResolution).toBe(POSENET_DEFAULTS.inputResolution);
-    expect(net.multiplier).toBe(POSENET_DEFAULTS.multiplier);
-    expect(net.quantBytes).toBe(POSENET_DEFAULTS.quantBytes);
+  it("instantiates bodyPose", () => {
+    expect(myBodyPose).toBeDefined()
+    expect(myBodyPose.model).toBeDefined();
   });
 
   it("detects poses in image", async () => {
-    const image = await asyncLoadImage(POSENET_IMG);
 
-    // Result should be an array with a single object containing pose and skeleton.
-    const result = await net.singlePose(image);
+    // Result should be an array with a single object containing the detection.
+    const result = await myBodyPose.detect(image);
     expect(result).toHaveLength(1);
-    expect(result[0]).toHaveProperty("pose");
-    expect(result[0]).toHaveProperty("skeleton");
+    expect(result[0]).toHaveProperty("box");
+    expect(result[0]).toHaveProperty("score");
+    expect(result[0].keypoints.length).toBeGreaterThanOrEqual(5);
 
     // Verify a known outcome.
-    const nose = result[0].pose.keypoints.find(
-      (keypoint) => keypoint.part === "nose"
+    const nose = result[0].keypoints.find(
+      (keypoint) => keypoint.name === "nose"
     );
+    // Should be {"name": "nose", "score": 0.7217329144477844, "x": 454.1112813949585, "y": 256.606980448618}
     expect(nose).toBeTruthy();
-    expect(nose.position.x).toBeCloseTo(448.6, 0);
-    expect(nose.position.y).toBeCloseTo(255.9, 0);
-    expect(nose.score).toBeCloseTo(0.999);
+    expect(nose.x).toBeCloseTo(454.1, 0);
+    expect(nose.y).toBeCloseTo(256.6, 0);
+    expect(nose.score).toBeCloseTo(0.721, 2);
+  });
+
+  it("calls the user's callback",(done) => {
+    expect.assertions(1);
+    const callback = (result) => {
+      expect(result).toHaveLength(1); // don't need to repeat the rest
+      done();
+    }
+    myBodyPose.detect(image, callback);
   });
 });

src/HandPose/index.test.js

@@ -3,6 +3,7 @@
 // This software is released under the MIT License.
 // https://opensource.org/licenses/MIT
 
+import crossFetch from 'cross-fetch';
 import { asyncLoadImage } from "../utils/testingUtils";
 import handpose from "./index";
 
@@ -14,10 +15,17 @@ describe("Handpose", () => {
 
   beforeAll(async () => {
     jest.setTimeout(10000);
-    handposeInstance = await handpose();
+
+    // TODO: this should not be necessary! Should already be handled by setupTests.js.
+    if (!global.fetch) {
+      global.fetch = crossFetch;
+    }
+
+    handposeInstance = handpose();
+    await handposeInstance.ready;
   });
 
-  it("detects poses in image", async () => {
+  it("detects hands in image", async () => {
     testImage = await asyncLoadImage(HANDPOSE_IMG);
     const handPredictions = await handposeInstance.predict(testImage);
     expect(handPredictions).not.toHaveLength(0);

src/utils/testingUtils/index.js

@@ -0,0 +1,36 @@
+// eslint-disable-next-line import/no-extraneous-dependencies
+import { createImageData, ImageData } from "canvas";
+
+export const asyncLoadImage = async (src) => {
+  const img = new Image();
+  if (src.startsWith("http")) {
+    img.crossOrigin = "true";
+  }
+  img.src = src;
+  await new Promise((resolve) => {
+    img.onload = resolve;
+  });
+  return img;
+};
+
+export const getRobin = async () => {
+  return asyncLoadImage(
+    "https://cdn.jsdelivr.net/gh/ml5js/ml5-library@main/assets/bird.jpg"
+  );
+};
+
+export const randomImageData = (width = 200, height = 100) => {
+  const length = width * height * 4; // 4 channels - RGBA
+  // Create an array of random pixel values
+  const array = Uint8ClampedArray.from({ length }, () =>
+    Math.floor(Math.random() * 256)
+  );
+  // Initialize a new ImageData object
+  return createImageData(array, width, height);
+};
+
+export const polyfillImageData = () => {
+  if (!global.ImageData) {
+    global.ImageData = ImageData;
+  }
+};