Classify Image (Triton)

How it fits

Typical backends

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.
• Hard constraintThe preprocessing (input size, color layout, channel order, mean/std, rescale, batch dimension, and whether outputs are logits) must match exactly what the served model expects, or the predictions will be meaningless even though the graph runs.
• Hard constraintInference runs against a Triton service that must be provisioned with the model; without it the component cannot score frames.
• 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.
• AccuracyClass labels and score calibration are defined by the served model, not by this component.