LeNet-5 was the convolutional neural network architecture introduced by Yann LeCun and colleagues in 1998.
The network was designed for recognizing handwritten digits and other visual patterns. Its structure combined convolutional layers, subsampling layers, and fully connected layers, an arrangement that became influential in later computer-vision systems.
LeNet-5 was applied to practical document-processing tasks, including reading handwritten numbers on checks. Its success showed that neural networks could learn useful visual features directly from pixel data rather than relying entirely on hand-designed rules.
LeNet-5 is sometimes confused with AlexNet, which was introduced much later and achieved a major ImageNet result in 2012. LeNet-5 was smaller and targeted at relatively simple grayscale character images, while later networks handled much larger and more complex natural-image datasets.