TensorFlow.js – Now Build Machine Learning Models in JavaScript!

Pranav Dar 02 Apr, 2018 • 2 min read

Overview

  • TensorFlow.js is an open source library that lets you build and train ML models in your browser
  • It’s available with GPU acceleration and also automaticlaly supports WebGL
  • You can import existing pre-trained models and also re-train entire existing ML models in the browser itself

 

Introduction

If you’re a JavaScript developers who’s new to the world of Machine Learning, or a Machine Learning practitioner who’s new to JavaScript, this latest release of TensorFlow will intrigue you.

TensorFlow.js is an open-source library that you can use to train and build machine learning models in your web browser, using JavaScript and APIs. If you’re familiar with Keras, the high level layers API will seem very familiar to you.

You might be wondering at this point what’s the advantage here of using a browser to build ML models.

Well, running a ML program in your browser means there is need to install any libraries or drivers! Open your browser page and you’re ready to build your model. It’s also available with GPU acceleration. Additionally, TensorFlow.js supports WebGL, a JavaScript API that is used for rendering 2D and 3D graphics within any browser.

You can also open the webpage on your mobile or tablet which will enable your model to take advantage of sensor data.

You can import an existing pre-trained model and TensorFlow.js converters will make it browser ready for you. You can also re-train existing ML models using “transfer learning to augment an existing model trained offline using a small amount of data collected in the browser using a technique called Image Retraining”. You can re-train the model very quickly and efficiently with this while only requiring a small amount of data.

 

How do I install TensorFlow.js?

It’s pretty straightforward. You can use it by installing in from NPM:

yarn add @tensorflow/tfjs
npm install @tensorflow/tfjs
Or from script:
<script src="https://cdn.jsdelivr.net/npm/@tensorflow/[email protected]"></script>

Check out the TensorFlow.js website and their GitHub page to read more about this release.

 

Our take on this

This is awesome news for folks who are familiar with JavaScript and are trying to carve their way in the ML world! It makes things a lot simpler for folks coming from a non-ML background who are looking to understand this field. The use cases for this are plenty. I highly encourage you to check out the demos on the official site which include a real-time performance on a piano by a neural network and playing the popular Pac-Man game using images trained in a browser.

TensorFlow.js is basically the successor of deeplearn.js. The major difference between the two is the TensorFlow.js includes a layers API and imports pre-trained models and can also re-train them. Also, you can work on almost any GPU but it will not be close to the speed you’ll get on CUDA.

Are you planning to use this for building your models? What do you think about this latest TensorFlow release? Let us know in the comments section below!

 

Subscribe to AVBytes here to get regular data science, machine learning and AI updates in your inbox!

 

Pranav Dar 02 Apr 2018

Senior Editor at Analytics Vidhya. Data visualization practitioner who loves reading and delving deeper into the data science and machine learning arts. Always looking for new ways to improve processes using ML and AI.

Frequently Asked Questions

Lorem ipsum dolor sit amet, consectetur adipiscing elit,

Responses From Readers

Clear

The detectorist
The detectorist 14 May, 2018

Hi, I am looking for an eample of an implementation of a time series inside a tensorflow.js. I am able to create tensor1d, 2d, 3d... but I don't know how to add time series into a tensor. Can you please write an example on your blog if you have expertise?

Sankirtan
Sankirtan 13 Jul, 2018

Does tensorflow JS allows us to make models using CSV files?

Related Courses