datumaro.experimental.categories#

Categories definitions for the experimental dataset system.

This module provides category management functionality using standard dataclasses instead of attrs, taking inspiration from the original Categories implementation.

Classes

`Categories`()	A base class for annotation metainfo.
`Colormap`(_data, ...)	A colormap that stores index-to-color mappings and provides efficient reverse lookup via an inverse colormap property.
`GroupType`(value[, names, module, qualname, ...])	Types of label groups for organizing labels.
`LabelCategories`(labels, group_type)	Represents a group of labels with a specific group type.
`MaskCategories`(colormap)	Describes a color map for segmentation masks.
`RgbColor`(r, g, b)	RGB color representation with named fields.

class datumaro.experimental.categories.GroupType(value, names=None, *, module=None, qualname=None, type=None, start=1, boundary=None)[source]#

Bases: IntEnum

Types of label groups for organizing labels.

EXCLUSIVE = 0#

INCLUSIVE = 1#

to_str() → str[source]#

classmethod from_str(text: str) → GroupType[source]#

class datumaro.experimental.categories.Categories[source]#

Bases: object

A base class for annotation metainfo. It is supposed to include dataset-wide metainfo like available labels, label colors, label attributes etc.

class datumaro.experimental.categories.LabelCategories(labels: ~typing.List[str] = <factory>, group_type: ~datumaro.experimental.categories.GroupType = GroupType.EXCLUSIVE)[source]#

Bases: Categories

Represents a group of labels with a specific group type.

labels: List[str]#

group_type: GroupType = 0#

add(label: str) → int[source]#

Add a new label.

Parameters:: label – The label name
Returns:: The index of the newly added category
Raises:: ValueError – If label already exists

find(name: str) → Tuple[int | None, str | None][source]#

Find a label by name.

Parameters:: name – The label name to find
Returns:: A tuple of (index, category) or (None, None) if not found

class datumaro.experimental.categories.RgbColor(r: int, g: int, b: int)[source]#

Bases: NamedTuple

RGB color representation with named fields.

Create new instance of RgbColor(r, g, b)

r: int#: Alias for field number 0

g: int#: Alias for field number 1

b: int#: Alias for field number 2

class datumaro.experimental.categories.Colormap(_data: ~typing.Dict[int, ~datumaro.experimental.categories.RgbColor] = <factory>)[source]#

Bases: object

A colormap that stores index-to-color mappings and provides efficient reverse lookup via an inverse colormap property.

property inverse_colormap: Dict[RgbColor, int]#: Get the inverse colormap (color -> index mapping).

get(index: int, default=None)[source]#: Get color by index with default.

class datumaro.experimental.categories.MaskCategories(colormap: ~datumaro.experimental.categories.Colormap = <factory>)[source]#

Bases: Categories

Describes a color map for segmentation masks.

colormap: Colormap#

classmethod generate(size: int = 255, include_background: bool = True) → MaskCategories[source]#

Generates MaskCategories with the specified size.

If include_background is True, the result will include the item: “0: (0, 0, 0)”, which is typically used as a background color.