This block is used in Convolutional Neural Networks; it represents a padding operation, which means adding zeros “around” the input to get a larger tensor.
This block is typically placed before a 2D Convolutional layer.
2D Convolutional layers, return by default a smaller image than the input. The output size is given by:
where: o = output size, i = input size, k = filter (kernel) size, s = stride, p = padding size.
The 2D Zero padding block allows you to manually specify the amount of horizontal and vertical padding you would like to add to your input.
Horizontal padding: The amount of zero-filled columns to add to the left and right of the input. Default: 1
Vertical padding: The amount of zero-filled rows to add to the left and right of the input. Default: 1