added tf.ready()

Author: ziyuan-linn

src/Handpose/index.js

@@ -44,7 +44,7 @@ class Handpose extends EventEmitter {
       modelType: this.config?.modelType ?? "full", // use full version of the model by default
       solutionPath: "https://cdn.jsdelivr.net/npm/@mediapipe/hands", // fetch model from mediapipe server
     };
-
+    await tf.ready();
     this.model = await handPoseDetection.createDetector(pipeline, modelConfig);
 
     this.modelReady = true;

src/NeuralNetwork/index.js

@@ -113,6 +113,7 @@ class DiyNeuralNetwork {
    * @param {*} callback
    */
   init(callback) {
+    tf.setBackend("webgl");
     // check if the a static model should be built based on the inputs and output properties
     if (this.options.noTraining === true) {
       this.createLayersNoTraining();

src/PoseDetection/index.js

@@ -78,10 +78,9 @@ class PoseDetection extends EventEmitter {
           bodyPoseDetection.movenet.modelType.MULTIPOSE_LIGHTNING;
     }
     // Load the detector model
-    await tf.setBackend("webgl");
+    await tf.ready();
     this.model = await bodyPoseDetection.createDetector(pipeline, modelConfig);
     this.modelReady = true;
-
     if (this.video) {
       this.predict();
     }