ImageNet

A massive hierarchical image database containing over 14 million manually annotated objects used to train and benchmark computer vision systems.

Launched in 2009 by Fei-Fei Li at Stanford, ImageNet organized over 14 million images into 20,000 categories based on the WordNet hierarchy. The project provided the scale necessary for the 2012 breakthrough of AlexNet, which utilized deep convolutional neural networks to dominate the ImageNet Large Scale Visual Recognition Challenge (ILSVRC). By offering a standardized dataset for object detection and image classification, it shifted the AI field from algorithmic design toward data-driven deep learning. It remains the primary benchmark for evaluating the accuracy of modern visual recognition models.

https://www.image-net.org