Output of a convolutional layer

Question

Is the calculated output correct?

Answer 1

For a convolution with default "configuration", your result is correct.

Note that there are other factors defining the behaviour of a convolution apart from the input and kernel, like padding, stride, dilation, etc. For convolutions, typical defaults are padding = 0 and stride = 1. With those values, your result is correct. Here's the code I used to check it:

import torch
import torch.nn.functional as F
x = torch.tensor(
  [[[0, 0, 1, 0, 0],
    [0, 1, 1, 0, 0],
    [1, 0, 1, 0, 0],
    [0, 0, 1, 0, 0],
    [0, 0, 1, 0, 0]]])
kernel = torch.tensor(
  [[0, 1, 0],
   [0, 1, 0],
   [0, 1,0]]).unsqueeze(0).unsqueeze(0)
F.conv2d(x, kernel, padding=0, stride=1)

If we set other values for padding and stride, we obtain different results. For instance, with F.conv2d(x, kernel, padding=1, stride=2) we obtain:

[[[0, 2, 0],
  [1, 3, 0],
  [0, 2, 0]]]