DEV Community

Cover image for Image Classification - Machine Learning in JavaScript
Bibek
Bibek

Posted on • Edited on • Originally published at blog.bibekkakati.me

Image Classification - Machine Learning in JavaScript

#javascript #machinelearning #webdev #node

Hello readers, recently I came to know about an awesome machine learning library for Javascript.

Any guess? No, it is not tensorflow.js.

It is ml5.js. So what is ml5.js?

Before proceeding, I would like to mention that I am not a machine learning expert. I just have some basic knowledge and try out these kinds of stuff in my free time with the help of external libraries.

In this article, I will brief about the library and show some examples.

What is ml5.js?

ml5.js is a machine learning library built on top of tensorflow.js which we can use in our web browser. It is being developed to make machine learning more accessible and life easier for people who are new to the Machine Learning arena.

It uses the browser's GPU to run all the calculations. Getting started with the library is very easy.

Just include the package link in your project and you are good to go.

<script src="https://unpkg.com/ml5@latest/dist/ml5.min.js"></script>
Enter fullscreen mode Exit fullscreen mode

Example

Let's implement something to understand the library.

We will implement Image Classification using this library.

What is Image Classification?

Image Classification is a computer vision technique in which we classify images according to the visual content in it. We train the Image Classifier with crafted data so that it can predict what type of object is in an Image. If we train the classifier with dog's images, it will be able to identify a dog in a photo very easily.

We will create a webpage where user can upload an image or paste an image link, and our classifier will classify what type of object is in the image.

Prerequisite 

- HTML
- CSS
- Javascript
Enter fullscreen mode Exit fullscreen mode

Code Implementation

First, we will be implementing the index.html page, which is our main web page.
On this page, we will have two buttons:

  • Upload Image to upload an image from your local drive
  • Paste Link to use hosted image from external server

On uploading or pasting a link of an image, the image will get rendered on the screen. After that on clicking the button What is in the image?, the result will be shown below it.

Some external links of the images will not work because of cors policy.

HTML 

<!DOCTYPE html>
<html lang="en">

<head>
  <meta charset="UTF-8">
  <meta http-equiv="X-UA-Compatible" content="IE=edge">
  <meta name="viewport" content="width=device-width, initial-scale=1.0">
  <title>Image Classifier</title>
  <script src="https://unpkg.com/ml5@latest/dist/ml5.min.js"></script>
  <link rel="stylesheet" href="./style.css">
</head>

<body onload="main()">
  <div class="loader-view" id="loaderView">
    <h3>Initializing ...</h3>
  </div>
  <div class="main" id="mainView">
    <h1 class="heading">Image Classifier</h1>
    <div class="select-image">
      <input type="file" name="Image" id="selectImage" accept="jpg,jpeg,png" hidden>
      <button class="upload-button button" id="uploadButton">Upload Image</button>
      <button class="link-button button" id="linkButton">Paste Link</button>
    </div>
    <div class="image-view" id="imageViewContainer">
      <img src="" alt="" class="image" id="imageView" crossorigin="anonymous">

      <button class="button" id="classifyButton">What is in the image ?</button>
      <h2 class="result" id="result"></h2>
    </div>

  </div>

  <script src="./script.js"></script>
</body>

</html>
Enter fullscreen mode Exit fullscreen mode

CSS

Just some styling for our webpage.

body {
    background-color: #000;
    color: #f0f8ff;
}

.main,
.loader-view {
    display: flex;
    flex-direction: column;
    justify-content: center;
    align-items: center;
    text-align: center;
    height: 100%;
    width: 100%;
}

.main {
    display: none;
}

.button {
    border: none;
    font-size: 16px;
    font-weight: 600;
    padding: 10px 15px;
    border-radius: 8px;
}

.upload-button {
    color: #f0f8ff;
    background: #ff7b00;
}

.link-button {
    color: #1f1f1f;
    background: #fdf8f4;
}

.upload-button:hover {
    background: #f0f8ff;
    color: #ff7b00;
}

.link-button:hover {
    color: #ff7b00;
}

.image-view {
    margin-top: 50px;
    width: 50vw;
    height: 60vh;
    display: none;
    flex-direction: column;
}

.image {
    max-width: 100%;
    max-height: 100%;
    border: solid #f0f8ff 6px;
    border-radius: 4px;
    margin-bottom: 5px;
}

.result {
    text-transform: capitalize;
    letter-spacing: 0.5px;
}
Enter fullscreen mode Exit fullscreen mode

Javascript

Calling the main function from index.html when body loads.

function main() {
    const classifier = ml5.imageClassifier("MobileNet", modelLoaded);
    const uploadButton = document.getElementById("uploadButton");
    const linkButton = document.getElementById("linkButton");
    const classifyButton = document.getElementById("classifyButton");
    const selectImage = document.getElementById("selectImage");
    const imageViewContainer = document.getElementById("imageViewContainer");
    const imageView = document.getElementById("imageView");
    const result = document.getElementById("result");
    const loaderView = document.getElementById("loaderView");
    const mainView = document.getElementById("mainView");

    uploadButton.onclick = function (e) {
        selectImage.click();
    };

    classifyButton.onclick = function (e) {
        classify(imageView);
    };

    linkButton.onclick = function (e) {
        const link = prompt("Paste Image Link Here");
        if (link != null && link != undefined) {
            imageView.src = link;
            imageViewContainer.style.display = "flex";
            result.innerText = "";
        }
    };

    selectImage.onchange = function () {
        if (this.files && this.files[0]) {
            var reader = new FileReader();
            reader.onload = function (e) {
                imageView.src = e.target.result;
                imageViewContainer.style.display = "flex";
                result.innerText = "";
            };
            reader.readAsDataURL(this.files[0]);
        }
    };

    function modelLoaded() {
        loaderView.style.display = "none";
        mainView.style.display = "flex";
    }

    function classify(img) {
        classifier.predict(img, function (err, results) {
            if (err) {
                return alert(err);
            } else {
                result.innerText = results[0].label;
            }
        });
    }
}
Enter fullscreen mode Exit fullscreen mode

Javascript Explanation 

const classifier = ml5.imageClassifier("MobileNet", modelLoaded);
Enter fullscreen mode Exit fullscreen mode

ml5.imageClassifier method is called to initialize the machine learning model.
Passing two arguments to the method first is the model which is MobileNet and a callback modelLoaded which will get invoked once the initialization is completed.

MobileNet is a term that describes a type of machine learning model architecture that has been optimized to run on platforms with limited computational power, such as applications on mobile or embedded devices. MobileNets have several use cases including image classification, object detection, and image segmentation. This particular MobileNet model was trained for Image Classification.

We have some other choices too for the model like:

  • Darknet
  • DoodleNet

You can use your own image classification model trained in Teachable Machine.

Then we are getting some references to buttons and views of our HTML file to listen for events and manipulate content and CSS styling.

uploadButton.onclick = function (e) {
    selectImage.click();
};
Enter fullscreen mode Exit fullscreen mode

On clicking the Upload Image button, it will click the image selector input element selectImage.

selectImage.onchange = function () {
    if (this.files && this.files[0]) {
        var reader = new FileReader();
        reader.onload = function (e) {
            imageView.src = e.target.result;
            imageViewContainer.style.display = "flex";
            result.innerText = "";
        };
        reader.readAsDataURL(this.files[0]);
    }
};
Enter fullscreen mode Exit fullscreen mode

When the user selects the image, selectImage.onchange listener will get invoked and it will set the image in imageView src.

linkButton.onclick = function (e) {
    const link = prompt("Paste Image Link Here");
    if (link != null && link != undefined) {
        imageView.src = link;
        imageViewContainer.style.display = "flex";
        result.innerText = "";
    }
};
Enter fullscreen mode Exit fullscreen mode

On clicking the Paste Link button, linkButton.onclick listener will get invoked and it will ask the user for the image link and if a link is provided, it will set the link in the imageView src.

classifyButton.onclick = function (e) {
    classify(imageView);
};
Enter fullscreen mode Exit fullscreen mode

On clicking the What is in the image? button, classifyButton.onclick listener will get invoked and it will call the classify method, which is responsible for image classification. Will pass the image element reference i.e, imageView to the classify method.

function modelLoaded() {
    loaderView.style.display = "none";
    mainView.style.display = "flex";
}
Enter fullscreen mode Exit fullscreen mode

This method will get invoked when our model is initialized and it is manipulating some CSS style to hide the loader.

function classify(img) {
    classifier.predict(img, function (err, results) {
        if (err) {
            return alert(err);
        } else {
            result.innerText = results[0].label;
        }
    });
}
Enter fullscreen mode Exit fullscreen mode

This method is the important method that is calling the predict method of classifier object. The predict method expects two arguments:

  • input which is a reference to the image element
  • callback function to handle results and error

On error, we are throwing it in an alert box.

Accessing the result from the results array, which contains multiple objects with label and confidence level.

Result

imgclassifier.png

You can find the files here: GitHub Repo

Try it out: Live Link

Originally published on blog.bibekkakati.me

Enjoyed? Give it a thumbs-up.

Thank you for reading | Feel free to connect 👋

Top comments (3)

Collapse
 
polaroidkidd profile image
Daniel Einars •

Just FYI, you can add syntax highlighting to your code comments by specifying the language after the first three backticks.

CSS:


html {
  background: red;
}
Enter fullscreen mode Exit fullscreen mode

JS


const myObjFn = (param) => {
  console.log(param)
}
Enter fullscreen mode Exit fullscreen mode
Collapse
 
bibekkakati profile image
Bibek •

Hey Daniel, Thank you for pointing that out.

Collapse
 
bibekkakati profile image
Bibek •

Hey @thatanjan , Thank you! I really appreciate your feedback.

Read next

midhunkrishnan profile image

To access terminal profile of VS Code and add commands to be run initially

Midhun Krishnan R -

gittech profile image

Github's Top 36 items of Dec 19, 2024

Gittech -

bharath_vasishtha profile image

Understanding Java Strings: A Comprehensive Guide

Bharath S R -

turingvangisms profile image

Tutorial: Laravel Next.js Tutorial

Turing -