Stride & Padding

What is Stride & Padding?

Two key hyperparameters that control the output dimensions of a convolutional layer. 'Stride' is the step size the filter moves across the input image (e.g., a stride of 2 downsamples the input by half). 'Padding' adds border pixels (usually zeros) to the input to preserve spatial dimensions or ensure the filter fits correctly. Together, they determine the size of the feature maps in a CNN.

Where did the term "Stride & Padding" come from?

Fundamental concepts in the design of Convolutional Neural Networks (CNNs).

How is "Stride & Padding" used today?

Ubiquitous in all computer vision architectures (ResNet, VGG, EfficientNet) to manage spatial resolution.

Related Terms