React and Tensorflow.js : Quickstart

React is a fantastic framework for modularizing Javascript projects and here I’m going to give a quick run down on how to start using Tensorflow.js within a React project.

All the code for this can be found here but the main secret is just to import TensorFlow :

import * as tf from '@tensorflow/tfjs';

From here we can start creating and manipulating tensors with all the utilities given to us by the TensorFlow team, see their documentation

const rotate90 = tf.tensor([[0, -1], [1, 0]])

You might recognize this as a matrix which rotates a vector by 90 degrees, let's see if it works

const a = tf.tensor([0, 1])

const b = tf.dot(a, rotate90)

And we get the expected result

 console.log(` result ${b.print()}`)
// Tensor [1, 0]

I’m pretty excited by this, software is eating the world and linear algebra is eating software so I suspect there’ll be a lot of excuses to use it on the front end in the future