Convolutional Neural Network

The Convolutional Neural Network (CNN) is the deep learning architecture that uses shared-weight filters (kernels) to efficiently process 3D data volumes (e.g., a 32x32x3 image) for superior image recognition and computer vision tasks.

A CNN is a specialized deep learning model, structured as a sequence of three main layer types: Convolutional, Pooling, and Fully-Connected. The core Convolutional layer applies small, learnable filters (kernels, e.g., 5x5x3) across the entire input volume to generate feature maps. This weight-sharing design significantly reduces the parameter count, making the model scalable for high-resolution images. CNNs became the de-facto standard for computer vision following their breakthrough performance in the 2012 ImageNet Large Scale Visual Recognition Challenge (ILSVRC), powering everything from facial recognition to autonomous vehicle perception.

https://cs231n.github.io/convolutional-networks/