Issue when I serve the models from my local server

[jhayush0502]

Error ->

extractWeightEntryFactory.ts:8 Uncaught (in promise) Error: expected weightMap[dense0/conv3/depthwise_filter] to be a Tensor4D, instead have undefined
    at extractWeightEntryFactory.ts:8:1
    at extractSeparableConvParamsFactory.ts:33:1
    at i (loadParamsFactory.ts:27:1)
    at d$ (extractParamsFromWeightMap.ts:17:1)
    at Np.extractParamsFromWeightMap (FaceFeatureExtractor.ts:47:1)
    at Np.loadFromWeightMap (NeuralNetwork.ts:113:1)
    at Ep.extractParamsFromWeightMap (FaceProcessor.ts:66:1)
    at Ep.loadFromWeightMap (NeuralNetwork.ts:113:1)
    at Ep.loadFromUri (NeuralNetwork.ts:96:1)
    at async Promise.all (:3000/index 1)

loading into React App ->

useEffect(() => {
  const loadModels = async () => {
    const MODEL_URL1 = process.env.PUBLIC_URL + '/models'
    console.log(MODEL_URL1)
    const MODEL_URL = 'http://localhost:3001/models'
    console.log(MODEL_URL)

    setInitialization(true)
    await Promise.all([
      faceapi.nets.tinyFaceDetector.loadFromUri(`${MODEL_URL}/tiny_face_detector_model-weights_manifest.json`),
      faceapi.nets.faceLandmark68Net.loadFromUri(`${MODEL_URL}/face_landmark_68_tiny_model-weights_manifest.json`),
      faceapi.nets.faceRecognitionNet.loadFromUri(`${MODEL_URL}/face_recognition_model-weights_manifest.json`),
      faceapi.nets.faceExpressionNet.loadFromUri(`${MODEL_URL}/face_expression_model-weights_manifest.json`)
    ]).then(startVideo)
  }
  loadModels()
}, [])

In my local express server is serving files at port 3001. At the root I have folder name 'models' which contain all the models loaded manifest.json + shard file .

How can I resolve this issue and serve models from local server.

[vladmandic]

load method takes uri of a folder containing model json, not json itself.

  const modelPath = 'http://localhost:3001/models'
  await faceapi.nets.tinyFaceDetector.load(modelPath);
  await faceapi.nets.ssdMobilenetv1.load(modelPath);
  await faceapi.nets.ageGenderNet.load(modelPath);
  await faceapi.nets.faceLandmark68Net.load(modelPath);
  await faceapi.nets.faceRecognitionNet.load(modelPath);
  await faceapi.nets.faceExpressionNet.load(modelPath);

and if you're using faceapi on the SAME express server, then why specify FQDN to start with, its much better to use relative path

  const modelPath = '/models'

[jhayush0502]

Can you please help me in one more model loading...

I am using tensorflow/handpose model and I had downloaded the model.json,anchors.json and shard files of it. But at the initial run it is loading bundle.js which require internet connection. How can I also serve that bundle.js through local server ...

const runHandpose = async () => {
const model1Url = 'http://localhost:3001/handModelDetector';
const model2Url = 'http://localhost:3001/handModelSkeleton';

const net = await handpose.load({
  model: {
    modelUrl: [`${model1Url}/model.json,${model2Url}/model.json`],
    anchorUrl: `${model1Url}/anchor.json`,
    shardUrls: [
      `${model1Url}/group1-shard1of2.bin`,
      `${model1Url}/group1-shard2of2.bin`,
      `${model2Url}/group1-shard1of2.bin`,
      `${model2Url}/group1-shard2of2.bin`,
    ],
    bundleUrl: `${model1Url}/bundle.js`,
  },
});
   // const net = await handpose.load()
    console.log('Handpose model loaded.')
    setInterval(() => {
      //for detecting hand in net
      detect(net)
    }, 100)
  }

[vladmandic]

i have no idea what bundle.js is, its not part of face-api and not something face-api needs at all. face-api is a single library.

[jhayush0502]

Yaa sir, it is related to tensorflow/handpose model.
Thanks for the response

[jhayush0502]

Thanks Alot @vladmandic It get resolved