ImageNet
Dataset
Vision
The 14-million-image labeled dataset that started the deep-learning era.
You're looking at a dataset record: a training-data collection — what's in it and which models learned from it.
Fei-Fei Li's ImageNet (2009) bet that data, not algorithms, was the bottleneck — and its annual competition proved it when AlexNet's 2012 GPU-trained entry crushed the field. The template it set — big labeled data plus benchmark competition — built modern AI's culture.
How this connects to the rest of the stack
This record's links at a glance — click a group to jump to it.
Used to train
Related
Sources
Record reviewed as of 2026-01-15 — see Methodology.