AlexNet won the 2012 ImageNet competition and helped trigger the modern deep-learning boom.
AlexNet was designed by Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton. It was a deep convolutional neural network trained on graphics processing units, or GPUs. Its architecture used rectified linear units, dropout, and data augmentation, techniques that helped it train effectively on a very large image dataset.
At the 2012 ImageNet Large Scale Visual Recognition Challenge, AlexNet achieved a top-five error rate of 15.3%, far ahead of the next entry’s result of about 26.2%. That performance showed that large neural networks, substantial labeled datasets, and GPU computation could dramatically improve image recognition. AlexNet was not the first convolutional neural network, but its competition result made the approach influential across computer vision and later AI research.