无法在Tensorflow.js中加载冻结的模型

Question

I try to load converted tensorflow model, but console shows the error as the following:

Version:

"dependencies": {
    "@tensorflow-models/posenet": "0.1.2",
    "@tensorflow/tfjs": "0.11.4",
    "@tensorflow/tfjs-converter": "^0.6.1",
    "asyncawait": "^1.0.7",
    "stats.js": "^0.17.0"
  }

Code:

import * as tf from '@tensorflow/tfjs';
import {loadFrozenModel} from '@tensorflow/tfjs-converter';

const MODEL_URL = 'faster_rcnn/web_model/tensorflowjs_model.pb';
const WEIGHTS_URL = 'faster_rcnn/web_model/weights_manifest.json';

const model =(async () => {
    await loadFrozenModel(MODEL_URL, WEIGHTS_URL)})();

Error:

I put the model in Google drive:

https://drive.google.com/file/d/1yNIC9TD1MWzJyI3ewnJCmM1bmTAYC_1E/view?usp=sharing

Please help.

Thank you!

Answer 1

You have to pass the storage type to loadFrozenModel() . Which would be in your case: file:/// .

Another problem in you code is that the immediatly called async function doesn't return anything, so your model variable stays undefined.

To fix both:

(async () => {
    const model = loadFrozenModel('file:///' + MODEL_URL, 'file:///' + WEIGHTS_URL);
    //Now you can only use 'model' in here.
})();