ImageNet is a huge collection of human-labelled photos, organised by word meanings (mostly nouns) from WordNet, that also became a well-known benchmark for image recognition AI.
In plain words
ImageNet groups pictures under concepts taken from WordNet, a dictionary-like database of word meanings, aiming for about 1,000 checked images per concept. The project was first described at the CVPR conference in 2009. From 2010 it ran a yearly contest, known as ILSVRC, where teams competed to classify and detect objects. In the 2012 contest, SuperVision, a deep convolutional network from the University of Toronto, had a top-5 error of about 15%, far below the next team's 26%.
A quick example
In the 2012 classification task, a model sees a photo and lists up to five of 1,000 possible categories, most confident first. It counts as right if the correct label is somewhere in that list.
Where you'll meet it
- Benchmarking and pre-training image recognition models.
- The yearly ILSVRC competition, run from 2010.
Related
Sources
- DocsAbout ImageNet, ImageNet · read 29 Sept 2026
- PaperImageNet Large Scale Visual Recognition Challenge, Russakovsky et al. (arXiv) · read 29 Sept 2026
- DocsILSVRC2012 results, ImageNet · read 29 Sept 2026
- DocsImageNet Large Scale Visual Recognition Challenge 2012 (ILSVRC2012), ImageNet · read 30 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.