Convolutional Neural Networks
A convolutional neural network (CNN) is a type of neural network designed for grid-like data such as images. It scans small filters across the input to detect local patterns like edges and textures, building on the assumption that pixels near each other are more related than pixels far apart.
Each convolutional layer slides a set of learned filters over the image and records where each pattern appears. Early layers pick up simple features such as edges and corners; deeper layers combine them into shapes, parts, and whole objects. Pooling layers shrink the result so the network can recognize a feature regardless of its exact position.
CNNs are a long-standing foundation of vision AI. On a production line, a CNN can classify each product image as good or defective, or locate foreign material on a conveyor. In public-sector use, the same approach can detect vehicles, cones, or standing water in camera feeds.
A CNN is one kind of neural network, not a synonym for it. CNNs are often compared with vision transformers, which use attention instead of local filters to relate distant parts of an image.