AI Glossary Term

What does “Convolutional Neural Network (CNN)” mean?

Definition

A neural architecture optimized for processing grid-like data such as images.

Browse the full AI GlossaryExplore all free AI guides

How Convolutional Neural Network (CNN) works

Diagram for Convolutional Neural Network (CNN), steps in order: Image pixels; then Convolution layers: detect edges; then Pooling: shrink and summarize; then Deeper layers: shapes, parts; then Final layer: class label.Image pixelsConvolution layers: detect edgesPooling: shrink and summarizeDeeper layers: shapes, partsFinal layer: class label

Example

A CNN that reads handwritten digits slides small filters across the image. Early filters pick up strokes and edges, later layers combine them into loops and lines, and the last layer says which digit it most likely is.

Why it matters

CNNs made image recognition practical and powered the deep learning boom around 2012. They are still used widely, though transformers now compete in many vision tasks.

Common questions

Why are CNNs good for images?

They reuse the same small filter across the whole image, so they find a pattern wherever it appears and need far fewer weights than a fully connected network.

What does 'convolution' mean here?

Sliding a small grid of weights over the input and computing a weighted sum at each position to detect a local pattern.

Are CNNs only for images?

Mostly, but they are also used on audio spectrograms, text, and other grid-like or sequential data.

Related terms

CLIPA multimodal model architecture that learns shared representations between text and images.Big DataVery large and complex datasets that require scalable storage and processing techniques.Diffusion ModelA generative architecture that learns to reverse noise to synthesize images, audio, or other content.EmbeddingA numeric vector representation that captures semantic meaning of text, images, or other data.Generative AIAI systems that produce new content such as text, images, audio, video, or code.Multimodal ModelA model that can process or generate multiple data types such as text, image, and audio.

Learn more in our free guides

Neural NetworksHard Parameter Sharing in Multi-Task NetworksBidirectional Recurrent NetworksSqueeze-and-Excitation Networks

Test yourself

Neural Networks QuizHard Parameter Sharing in Multi-Task Networks QuizBidirectional Recurrent Networks Quiz

See also

Continual LearningCross-Entropy LossContext WindowData AugmentationComputer VisionData DriftComputeData Labeling