Padding

Adding pixels around an image to control output size and keep edge information.

Padding adds extra pixels around the edges of an image before a convolution. It prevents the output from shrinking, makes sure filters cover edge and corner pixels so edge features are preserved, and gives precise control over the output size.

Example: a 5×5 image with a 3×3 filter and no padding gives a 3×3 output; with one pixel of zero padding the output stays 5×5.

output size = ⌊ \frac{n + 2 p - k}{s} ⌋ + 1

Formula to remember

Sigmoid / tanh / ReLU

σ (z) = \frac{1}{1 + e ^{- z}}, tanh (z), ReLU (z) = max (0, z)

Ranges (0,1), (−1,1) and [0,∞).

Quick check · +5 XP

Global average pooling on the 4×4 matrix with values 1 to 16 gives…

Practice this topic

Your notes

Saved in this browser as you type.