ImageNet
Large labelled image dataset and benchmark (mid-2000s onward) that made progress in computer vision measurable and paved the way for deep learning.
ImageNet is a large-scale image dataset organised into thousands of WordNet-inspired categories, begun in the mid-2000s under Fei-Fei Li and published as a database in 2009. Crowdsourced labelling produced millions of examples for supervised learning.
From 2010 it powered the ImageNet Large Scale Visual Recognition Challenge, a shared leaderboard that made vision progress comparable across labs. In 2012 AlexNet cut error rates sharply and is often dated as the start of the modern deep learning boom. The dataset itself showed that scale of labelled data can matter as much as the model. See the article and timeline for the full story.