Tensorflow.js tf.image.resizeNearestNeighbor() Function

Tensorflow.js is an open-source library developed by Google for running machine learning models as well as deep learning neural networks in the browser or node environment.

The .image.resizeNearestNeighbor() function is used to rescale a heap of 3D images to a different configuration.

Syntax:

tf.image.resizeNearestNeighbor(images, 
    size, alignCorners?, halfPixelCenters?)

Parameters:  

• images: The stated images of rank 4 or else rank 3, which is of configuration [batch, height, width, inChannels]. Moreover, in case its of rank 3, then the batch of one is presumed. It can be of type tf.Tensor3D, tf.Tensor4D, TypedArray, or Array.
• size: The different stated configuration [newHeight, newWidth] in order to rescale the images. Every channel is rescaled one by one. It is of type [number, number].
• alignCorners: It is the optional parameter whose by default value is false. In case its true, the input is resized by (new_height – 1) / (height – 1), that absolutely queues the four corners of the stated images as well as rescaled images. However, if its false then it is rescaled by new_height / height. It deals in the same manner with the width dimension. It is of type Boolean.
• halfPixelCenters: It is the optional parameter whose by default value is false. It is of type Boolean.

Return Value: It returns tf.Tensor3D, or tf.Tensor4D.

Example 1: In this example, we will be going to use a 4d tensor and a size parameter.

Javascript

// Importing the tensorflow.js library
import * as tf from "@tensorflow/tfjs"

// Calling image.resizeNearestNeighbor() method
// and printing output
tf.image.resizeNearestNeighbor(tf.tensor4d([[

  [[4, 7], [21, 9]],

  [[8, 9], [1, 33]]

]]), [3, 4]).print();

Output:

Tensor
    [[[[4 , 7 ],
       [4 , 7 ],
       [21, 9 ],
       [21, 9 ]],

      [[4 , 7 ],
       [4 , 7 ],
       [21, 9 ],
       [21, 9 ]],

      [[8 , 9 ],
       [8 , 9 ],
       [1 , 33],
       [1 , 33]]]]

Example 2: In this example, we will be going to use an array of floats, alignCorners, as well as halfPixelCenters.

Javascript

// Importing the tensorflow.js library
import * as tf from "@tensorflow/tfjs"

// Defining an array of floats
const arr = [[

  [[1.1, 1.7, 1.5, 1.1],
  [1.7, 1.9, 8.1, 6.3]],
  [[3.3, 3.4, 3.7, 4.0],
  [5.1, 5.2, 5.3, 5.9]]

]];

// Calling image.resizeNearestNeighbor() method
// and printing output
tf.image.resizeNearestNeighbor(arr, [1, 2], true, false).print();

Output:

Tensor
    [[[[1.1, 1.7, 1.5      , 1.1      ],
       [1.7, 1.9, 8.1000004, 6.3000002]]]]

Reference: https://js.tensorflow.org/api/latest/#image.resizeNearestNeighbor