Concept

Different types of convolutions

1- valid: when the padding is zero

2- same: when we choose padding so that output size of the image is the same as the input size.

3- full: similar to "same" padding but the filter's bottom right edge starts its placement from the top right corner of the input image i.e. enough zerosare added for every pixel to be visited "k" times. The figure below shows the same.

Image 0

0

2

Updated 2026-06-17

Tags

Data Science

Machine Learning Yearning @ DeepLearning.AI

Dive into Deep Learning @ D2L

Deep Learning

Machine Learning

Supervised Learning