← All models · BEiT

use case · basics

Basics: classify an image

The whole idea in its simplest form. Give BEiT a picture; the self-supervised transformer hands back its best guesses from the 1,000 ImageNet categories, each with a probability. Bigger bar = more sure.

Drop an image, or click to choose
Two cats on a couch A cozy bedroom A riverside at golden hour
Selected image

The bars are a softmax over all 1,000 classes, so the ones shown here rarely add to 100% — the rest of the probability is spread across the classes that didn't make the cut.

← Back to BEiT · Next: Practical →