Module configs 

The vocabulary types a model config is built from.

Per-role output descriptors (Detection, Boxes, Scores, Classes, Segmentation, Protos, MaskCoefficients, Mask), the axis names used to declare physical layout (DimName), quantization pairs (QuantTuple), and the enums that steer decoding: DecoderType, DecoderVersion, Nms, and the resolved ModelType.

Nms is worth a note: it has no None variant. Bypassing suppression is expressed as Option<Nms>::None on the decoder configuration, and Nms::Auto means “take the mode from the config document”, falling back to Nms::ClassAware when the document is silent.

Structs

Boxes

Classes

Detection

Mask

MaskCoefficients

Protos

QuantTuple

Scores

Segmentation

Enums

DecoderType

The post-processing family. The code is the one the C EF_DECODER_TYPE_* enumerators and the Python DecoderType enum use.

DecoderVersion

Decoder version for Ultralytics models.

DimName

A named tensor axis. The code is the one the C EF_DIM_NAME_* enumerators and the Python DimName enum use.

ModelType

Nms

NMS (Non-Maximum Suppression) mode for filtering overlapping detections.

OutputType

The kind of a model output, without its configuration: the tag of crate::ConfigOutput. The string form is that enum’s serialized type; the code is the one the C EF_OUTPUT_TYPE_* enumerators use.

Constants

NMS_OFF_CODE

The code for “NMS off” (Option::<Nms>::None). No Nms variant uses it.

Functions

deserialize_dshape

Deserialize dshape from either array-of-tuples or array-of-single-key-dicts.