TypeError: (0 , util_1.isNullOrUndefined) is not a function using @tensorflow/tfjs-node on mac

Question

I am trying to do some image recognition using @tensorflow/tfjs-node and mobilenet. Here is the code:

import fs from 'fs';
import * as tf from '@tensorflow/tfjs-node'; // TensorFlow.js for Node.js
import * as mobilenet from '@tensorflow-models/mobilenet'; // MobileNet for image classification
import sharp from 'sharp';

const classifyImage = async (imagePath) => {
  try {
    await tf.setBackend('tensorflow');

    // Resize the image to 224x224 as required by MobileNet
    const resizedImageBuffer = await sharp(imagePath)
      .resize(224, 224) // MobileNet requires 224x224 images
      .toBuffer();

    // Decode the resized image to a tensor
    const imageTensor = tf.node.decodeImage(resizedImageBuffer, 3);

    // Load the MobileNet model
    const model = await mobilenet.load();

    // Perform classification
    const predictions = await model.classify(imageTensor);

    console.log('Predictions:', predictions);

    // Clean up tensors to avoid memory leaks
    tf.dispose(imageTensor);
  } catch (error) {
    console.error('Error during image classification:', error);
  }
};

// Provide the image path as an argument
const imagePath = process.argv[2];
if (!imagePath) {
  console.error('Please provide an image path as the first argument.');
  process.exit(1);
}

classifyImage(imagePath);

program fails in line:

    const imageTensor = tf.node.decodeImage(resizedImageBuffer, 3);

with error:

node app.js test-image.jpg
(node:15224) [MODULE_TYPELESS_PACKAGE_JSON] Warning: Module type of file:///Users/albertmontolio/Documents/coding_area/interviews/koerber/image-recognition/app.js is not specified and it doesn't parse as CommonJS.
Reparsing as ES module because module syntax was detected. This incurs a performance overhead.
To eliminate this warning, add "type": "module" to /Users/albertmontolio/Documents/coding_area/interviews/koerber/image-recognition/package.json.
(Use `node --trace-warnings ...` to show where the warning was created)
Error during image classification: TypeError: (0 , util_1.isNullOrUndefined) is not a function
    at createTensorsTypeOpAttr (/Users/albertmontolio/Documents/coding_area/interviews/koerber/image-recognition/node_modules/@tensorflow/tfjs-node/dist/nodejs_kernel_backend.js:675:38)
    at Object.kernelFunc (/Users/albertmontolio/Documents/coding_area/interviews/koerber/image-recognition/node_modules/@tensorflow/tfjs-node/dist/kernels/Cast.js:30:65)
    at kernelFunc (/Users/albertmontolio/Documents/coding_area/interviews/koerber/image-recognition/node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js:4707:32)
    at /Users/albertmontolio/Documents/coding_area/interviews/koerber/image-recognition/node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js:4767:27
    at Engine.scopedRun (/Users/albertmontolio/Documents/coding_area/interviews/koerber/image-recognition/node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js:4572:23)
    at Engine.runKernelFunc (/Users/albertmontolio/Documents/coding_area/interviews/koerber/image-recognition/node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js:4763:14)
    at Engine.runKernel (/Users/albertmontolio/Documents/coding_area/interviews/koerber/image-recognition/node_modules/@tensorflow/tfjs-core/dist/tf-core.node.js:4636:21)
    at cast_ (/Users/albertmontolio/Documents/coding_area/interviews/koerber/image-recognition/node_modules/@tensorflow/tfjs/dist/tf.node.js:4084:19)
    at cast__op (/Users/albertmontolio/Documents/coding_area/interviews/koerber/image-recognition/node_modules/@tensorflow/tfjs/dist/tf.node.js:4027:28)
    at getGlobalTensorClass.toInt (/Users/albertmontolio/Documents/coding_area/interviews/koerber/image-recognition/node_modules/@tensorflow/tfjs/dist/tf.node.js:33989:12)

This is my package.json:

{
  "name": "image-recognition",
  "version": "1.0.0",
  "main": "index.js",
  "scripts": {
    "test": "echo \"Error: no test specified\" && exit 1"
  },
  "keywords": [],
  "author": "",
  "license": "ISC",
  "description": "",
  "dependencies": {
    "@tensorflow-models/mobilenet": "^2.1.1",
    "@tensorflow/tfjs-node": "^4.22.0",
    "sharp": "^0.33.5"
  }
}

I am using node version 23.3.0.

I've tried many many things without success: check image, change image, install new versions, old versions. I don't know what else to try. If this internal method is broken, what can I do?

Answer 1

I found out that version 4.22.0 of tensorflow is not compatible with node 23 due to util_1.isNullOrUndefined. This util and others were removed in node 23. Solution was to downgrade node to version 22.