How to check or print padding done in python (keras)?

While making a model, we use padding in Conv2D function and also Maxpooling function like :

Conv2D(filters=52,kernel_size=(3,3),padding='same')

Is there a means of check the padding done or printing its size or the padding itself so as to facilitate debugging ?

The exact way how the padding is computed can be found in the github code here. First the output width and height values are computed and then the padding:

For padding="SAME":

out_height = ceil(in_height / stride_height)
out_width  = ceil(in_width / stride_width)

if (in_height % strides[1] == 0):
    pad_along_height = max(filter_height - stride_height, 0)
else:
    pad_along_height = max(filter_height - (in_height % stride_height), 0)
if (in_width % strides[2] == 0):
    pad_along_width = max(filter_width - stride_width, 0)
else:
    pad_along_width = max(filter_width - (in_width % stride_width), 0)

You can also get an idea about the padding by using a convolution with a kernel that consists of ones:

import tensorflow as tf

dummy_in = tf.ones([1, 6, 6, 1])
kernel_size = 3
strides = 1
padding = "SAME"

conv = tf.keras.layers.Conv2D(1, 
                              kernel_size, 
                              strides=strides, 
                              padding=padding, 
                              use_bias=False,
                              kernel_initializer=tf.ones)
# build layer
conv(dummy_in)

# subtracting the number of pixels in kernel gives
# the number of pixels that were padded
kernel_size**2 - conv(dummy_in)[...,0]

This outputs

[[[5., 3., 3., 3., 3., 5.],
  [3., 0., 0., 0., 0., 3.],
  [3., 0., 0., 0., 0., 3.],
  [3., 0., 0., 0., 0., 3.],
  [3., 0., 0., 0., 0., 3.],
  [5., 3., 3., 3., 3., 5.]]]

where each entry corresponds to the number of pixels that were padded in a single convolutional window. As you can see, 5 pixels where added around the corners, and 3 pixels at the sides. In this case, the image was padded by one row/column of zeros at each side.