Convolutional Neural Network CNN

A network that slides small learned filters across an image, building up from edges to shapes to objects.

Computer Vision Intermediate 1 min read

Also known as: CNN, ConvNet

Definition

A CNN applies small learned filters across the whole input, so a feature detector that finds an edge works anywhere in the image rather than only where it was trained. Stacking these layers builds a hierarchy: edges, then textures, then parts, then objects.

The two properties that made CNNs work — weight sharing and locality — also make them dramatically more parameter-efficient than a fully connected network on image data. Vision transformers have overtaken them at large scale, but CNNs remain the sensible default when data or compute is limited.

How it relates to other terms

Broader term Neural Network A model built from layers of simple weighted units, trained by adjusting those weights until the output matche... Contrast with Transformer The neural network architecture behind almost all modern language models, built around attention instead of re...
Referenced by
Neural Network Broader term

Was this definition helpful?

Definitions are reviewed by our editorial team. Spotted a problem? Tell us.

Keep reading

Browse the glossary Suggest a correction