• 1.0) Project description
• 1.1) Requirements (For W10/11 OS)
• 1.2) Project execution
• 1.3) Possible solutions for library installation errors

• 2.0) Image Reading and Decoding Layer Model
• 2.1) MobileNet Model Architecture

• 3.0) Official documentation
• 3.1) Unofficial documentation
• 3.2) Video tutorials

This module implements the MobileNet convolutional neural network model for image classification using TensorFlow.js in Node.js. It allows efficient image classification with the pre-trained MobileNet model. The MobileNet architecture is optimized for mobile and embedded vision applications while maintaining high accuracy. This implementation includes:

• Integration with TensorFlow.js and Node.js for server-side image classification
• Pre-trained MobileNet model weights for fast inference without training
• Support for classifying common objects across 1000 categories
• Efficient processing pipeline for image loading and classification
• Example code showing how to use the model for real-world image classification tasks
• High performance with low computational requirements compared to larger CNN models

• Create a working environment through an IDE (Visual Studio Code recommended for best TensorFlow.js integration)
• Clone the Project (git clone https://github.com/andresWeitzel/Modulo_MobileNET_CNN_Tensorflow_NodeJs)
• IMPORTANT: For using tensorflow on w10/11 it is necessary to have the Desktop development with C++ add-on for visual studio installed. To install it follow these steps. In addition to having Python installed and configured its PATH.
• Inside the directory install all implemented plugins
  • npm update (Update npm repositories)
  • npm install -g npm@latest (Latest compatible version of npm)
  • npm install -S @tensorflow/tfjs(TensorFlow.js Core) - Core TensorFlow.js library for building and training neural networks
  • npm install -S @tensorflow/tfjs-node (TensorFlow.js Node.js extension) - Node.js binding providing CPU/GPU acceleration
  • npm install -S @tensorflow-models/mobilenet (MobileNet Model) - Pre-trained MobileNet model for image classification
• Verify installations:
  • Run node -v to confirm Node.js version
  • Run npm -v to verify npm version
  • Run python --version to check Python installation
• Test the setup:
  • Create a test file and import TensorFlow.js
  • Try loading the MobileNet model
  • If any errors occur, check the troubleshooting section

• tensorflow/tfjs#6700
• tensorflow/tfjs#6670
• https://github.com/nodejs/node-gyp#on-windows

//Imports
const tf = require('@tensorflow/tfjs');
const tfnode = require('@tensorflow/tfjs-node');
const fs = require('fs');

module.exports.exec = (path) => {
    const imageBuffer = fs.readFileSync(path);
    const tfimage = tfnode.node.decodeImage(imageBuffer);
    return tfimage;
}

1. Reading the image file from the specified path
2. Converting the image into a format that TensorFlow can process
3. Decoding the image buffer into a tensor

Key components:

• Uses fs.readFileSync() to read the image file into a buffer
• Uses tfnode.node.decodeImage() to convert the buffer into a TensorFlow tensor
• Returns the tensor for use in classification

//Imports
const read= require('../fileSystem/read');
const mobileNet = require('@tensorflow-models/mobilenet');

const classificator = async(path) =>{
    try {
        const img = read.exec(path);
        const model = await mobileNet.load();
        const predictions = await model.classify(img);
        console.log('Classification Results : ', predictions);
    } catch (error) {
        console.log(error);
    }
}

if (process.argv.length !== 3) throw new Error('Incorrect arguments!');
classificator(process.argv[2]);

• Images are stored in the src/img directory
• We select an image for the model to classify.

• In my case I have removed the image metadata and name so that the model classifies according to its confidence level.
• We execute the model from src/runners/classificator with node classificator.js ../img/test04.jpg
• Expected Output...

2022-11-20 20:09:04.459159: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in 
performance-critical operations:  AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
Classification Results :  [
  { className: 'espresso', probability: 0.9372739195823669 },        
  { className: 'cup', probability: 0.02711259014904499 },
  { className: 'mortar', probability: 0.008303580805659294 }
]

• High probability for espresso (espresso coffee). High Confidence Interval and Correct Classification.

• In my case I have removed the image metadata and name so that the model classifies according to its confidence level.
• We execute the model from src/runners/classificator with node classificator.js ../img/test05.jpg
• Expected Output...

2022-11-20 20:11:07.652384: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in 
performance-critical operations:  AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
Classification Results :  [
  {
    className: 'microwave, microwave oven',
    probability: 0.3682105839252472
  },
  { className: 'coffeepot', probability: 0.08597368746995926 },      
  { className: 'paper towel', probability: 0.08168061077594757 }     
]

• High probability for microwave (microwave oven). We can notice that the Model does not apply an acceptable confidence interval for several objects. The model managed to predict that there is a coffeepot (coffee maker) but with low probability. Low Confidence Interval and Correct Classification. For this case, a fine-tuning of model parameters should be applied.

• TensorFlow Core
• TensorFlow Api NodeJs
• MobileNET v1
• TensorFlow.js Models
• TensorFlow.js Examples
• TensorFlow.js Tutorials
• TensorFlow.js Guide
• MobileNets: Efficient Convolutional Neural Networks
• TensorFlow Hub - MobileNet v1

• Base Reference
• Base Repository
• TensorFlow.js Image Classification Tutorial
• Building a Node.js Image Recognition App
• Image Classification with TensorFlow.js
• MobileNet Image Classification with TensorFlow.js

• Playlist functionality test