Max Pool 2D

Max_Pool2d is an image processing layer, used for downscaling images while retaining feature data

The Max_Pool2d layer has no learnable perameters. It applies an opperation similar to a convolution, by sliding a window over images. The resulting downscaled images are made up of the max values from each window position.

Output Dimensions

The dimensions of the output image are calculated the same as for Conv2d Referenced Here

Bases: Module

Scales down an image by selecting the maximum pixel values

Parameters:

Name	Type	Description	Default
kernel_size	int or tuple	Shape of kernel (h, w). If integer then defaults to (a, a)	required
stride	int or tuple	Number of pixels to skip in each direction (h, w). If integer then defaults to (a, a)	required
padding	int or tuple	number of zeros to apply to the input image vertically and horizontally (h, w). If integer then defaults to (a, a)	required

Shapes

Input:
    (N, channels, in_height, in_width)

Output
    (N, channels, out_height, out_width)

Implementation

import neuraltoolkit as ntk
import numpy as np

sample_count = 100
channels = 3
height = 28
width = 28

x = ntk.Tensor(
    np.random.rand(sample_count, channels, height, width)
)

layer = ntk.Max_Pool2d(
    kernel_size=3,
    stride=2,
    padding=0
)

y = layer(x)