Detection¶

Point detection — cell / nucleus centroids — on a frozen-encoder token grid. The default path regresses a per-class peak heatmap with a neural decoder; the rows below are the substrate and component choices that path can swap in, with the walkthrough for each where one exists.

Method	Summary	Walkthrough
Neural decoder (default)	A lightweight conv decoder regresses a per-class peak heatmap; the `DetectionHead` reads points back out and scores F1@δ.	Dense prediction
Multi-encoder composite	Concatenate the dense outputs of several foundation models into one richer per-position vector before the decoder.	Composite
Decoder-free pixel classifier	A per-pixel classifier on the encoder’s own attention. Not yet wired for detection — peaks need the decoder’s spatial smoothing to stay separable.