Padding and stride in TensorFlow - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is padding in convolutional neural networks?

Padding means adding extra pixels (usually zeros) around the edges of an image before applying convolution. It helps keep the output size larger or the same as the input.

beginner

What does stride control in convolution?

Stride controls how many pixels the filter moves each step when sliding over the image. A stride of 1 moves one pixel at a time, stride 2 moves two pixels, making the output smaller.

intermediate

What is the difference between 'valid' and 'same' padding in TensorFlow?

'Valid' padding means no padding is added, so the output is smaller than the input. 'Same' padding adds zeros so the output size stays the same as the input size.

intermediate

How does increasing stride affect the output size?

Increasing stride makes the filter jump more pixels, so the output size becomes smaller because fewer positions are covered.

beginner

Why might you use padding in a convolutional layer?

Padding helps preserve important edge information and controls output size. Without padding, edges get ignored and output shrinks quickly.

What does 'same' padding do in a convolutional layer?

A Keeps output size the same as input size

B Removes pixels from the input

C Doubles the output size

D Adds random noise to input

'Same' padding adds zeros around the input so the output size matches the input size.

If stride is set to 2, what happens to the output size?

A Output size becomes zero

B Output size doubles

C Output size stays the same

D Output size halves approximately

Stride 2 moves the filter two pixels at a time, reducing output size roughly by half.

Which padding type means no padding is added?

A Same

B Full

C Valid

D Zero

'Valid' padding means no padding, so output is smaller than input.

Why do we add padding in convolutional layers?

A To keep edge information and control output size

B To increase the number of filters

C To speed up training

D To reduce model size

Padding preserves edges and helps keep output size as desired.

What happens if you use stride 1 and no padding?

A Output size is larger than input

B Output size is smaller than input

C Output size stays the same

D Model will not train

Without padding and stride 1, output shrinks because filter can't slide over edges fully.

Explain in your own words what padding and stride do in a convolutional neural network.

Think about how the filter moves and what happens at image edges.

Describe the difference between 'valid' and 'same' padding and how they affect the output size.

Consider how zeros around the image change the convolution result.