Classify Image (HuggingFace)

@huggingface1 VersionClassify

Classifies an Image with a HuggingFace, timm, or Optimum network and projects one internal ranked result into the connected output arm: top-1 Classification, ranked [Classification], a String label, and more.

How it fits

Typical backends

Classify frames from a file source and publish the ranked classes as JSON.

Classify a video stream and overlay the predicted label on each frame for visual review.

Caveats

I/O contractA single ranked list is built per frame and every output arm is a projection of it, so all arms agree class-for-class; the arm is fixed by which output type the downstream connection resolves, not by a parameter.
State lifetimeThe model id and several inference knobs are hot-swappable at runtime and reload the backend in place, so switching weights needs no restart.
FallbackA class with no recorded name falls back to its stringified class id, and a requested GPU collapses to CPU when no CUDA hardware is visible.
Parameter interactionThe confidence floor is applied before the top selection, so classes below it are dropped first and the kept-count then bounds the remaining ranked list.
AccuracyThe label vocabulary is whatever the chosen network was trained on, so swapping the model changes both the label set and the score calibration.

Versions

198c9905latestdefault
Automated release
6/25/2026