02 · Neural Networks & Model Architectures · Topic 03

What is a CNN?

A Convolutional Neural Network (CNN) is a neural-network architecture designed to detect spatial patterns, especially in images.

What is it?

A Convolutional Neural Network (CNN) uses small learned filters that move across an input to detect local patterns.

In an image, these patterns can include edges, textures, and shapes. The network combines them into information useful for a prediction.

Think of it like...

Imagine scanning an image with a small window.

Edges — where light and dark meet
Textures — repeated visual details
Shapes — combinations of simpler patterns

How it works

1 · InputReceive an imagePixels enter the network.

2 · Convolution filtersScan for patternsLearned filters look across small local regions.

3 · Feature mapsKeep useful signalsEach filter shows where its pattern appears.

4 · Deeper feature layersCombine patternsLater layers build richer visual features.

5 · PredictionReturn an answerThe model classifies or locates what it sees.

Real-world examples

Image classification

Identify whether an image contains a cat, vehicle, or another category.

Object detection

Find objects in an image and mark where they are.

Medical imaging

Help analyze scans for visual patterns that need clinical review.

What it is NOT

CNN ≠ Computer Vision

CNN: One architecture used for visual tasks.

Computer Vision: The broader field of interpreting visual information.

CNN ≠ Neural Network

CNN: A specialized type of neural network.

Neural Network: The broader family of learned architectures.

CNN ≠ Image Generator

CNN: Often recognizes or locates patterns.

Image Generator: Creates new images; modern vision models also widely use Transformers.

Remember this

A CNN learns local spatial patterns by applying learned filters across an input.

Video

No real video asset is available for this topic yet.