class
Detection
extends
TransformsPresetDetection(max_size: int = 1333, min_area: float = 1.0, min_visibility: float = 0.0, mean: tuple[float, ...] | None = None, std: tuple[float, ...] | None = None, interpolation: str | Interpolation = Interpolation.BILINEAR)Object-detection preset — coordinates ride with the image.
Pipeline: LongestMaxSize(max_size) → PadIfNeeded(max_size, max_size) → Normalize(mean, std) plus a
BboxParams(min_area, min_visibility) policy on the
enclosing Compose so out-of-frame boxes (and their
labels) drop automatically after the pipeline runs.
Parameters
max_sizeint= 1333Longest-side target after resize; canonical detection
recipes use
800 / 1333. PadIfNeeded then squares the
canvas to (max_size, max_size) so batching works.min_areafloat= 1.0Drop post-pipeline boxes whose absolute pixel area is below
this — see
BboxParams.min_visibilityfloat= 0.0Drop boxes whose visible fraction (area / original area)
falls below this.
meantuple of float= NonePer-channel normalization stats; default ImageNet.
stdtuple of float= NonePer-channel normalization stats; default ImageNet.
No description.