keras-js/src/layers/convolutional/UpSampling2D.js

import Layer from '../../engine/Layer'
import Tensor from '../../Tensor'
import ops from 'ndarray-ops'

/**
 * UpSampling2D layer class
 */
export default class UpSampling2D extends Layer {
  /**
   * Creates a UpSampling2D activation layer
   * @param {number} size - upsampling factor
   */
  constructor (size = [2, 2], attrs = {}) {
    super(attrs)
    const {
      dimOrdering = 'tf'
    } = attrs

    this.size = size
    this.dimOrdering = dimOrdering
  }

  /**
   * Method for layer computational logic
   * @param {Tensor} x
   * @returns {Tensor} x
   */
  call = x => {
    // convert to tf ordering
    if (this.dimOrdering === 'th') {
      x.tensor = x.tensor.transpose(1, 2, 0)
    }

    const inputShape = x.tensor.shape
    const outputShape = [
      inputShape[0] * this.size[0],
      inputShape[1] * this.size[1],
      inputShape[2]
    ]
    let y = new Tensor([], outputShape)
    for (let i = 0; i < this.size[0]; i++) {
      for (let j = 0; j < this.size[1]; j++) {
        ops.assign(
          y.tensor.lo(i, j, 0).step(this.size[0], this.size[1], 1),
          x.tensor
        )
      }
    }
    x.tensor = y.tensor

    // convert back to th ordering if necessary
    if (this.dimOrdering === 'th') {
      x.tensor = x.tensor.transpose(2, 0, 1)
    }

    return x
  }
}