Skip to content

semsegdatadescritption

semsegdatadescritption

Module for SemSegClassInfo and SemSegDataDescription

Classes

SemSegClassInfo dataclass

Class for (target or output) class information (name and color of mask image)

Parameters:

  • color (List[int]) –

    Color as list of three uint8 values (rgb) or -1 as any value. e.g. [-1, 255, -1] will look for 255 in the g-channel and ignore r- and b-channel.

  • name (str) –

    Name of the class

SemSegDataDescription dataclass

Bases: InputImageDataDescription, OutputImageDataDescription

DataDescription for SemSeg data

Parameters:

  • classes (List[SemSegClassInfo]) –

    List[SemSegClassInfo] Each entry is an initialized SemSegClassInfo, for which the name (str) and color (List[int]) must be present. The index in output array (third axis) is defined by the order of the classes list. Usually the color has to be a list of length 3 with uint8 values. The only exception is the usage of -1 as any value. e.g. [-1, 255, -1] will look for 255 in the g-channel and ignore r- and b-channel.

Functions
get_class_idx_lut 
get_class_idx_lut()

Returns the class index lut to transform the mask image

Source code in niceml/data/datadescriptions/semsegdatadescritption.py 
66
67
68
def get_class_idx_lut(self) -> np.ndarray:
    """Returns the class index lut to transform the mask image"""  # QUEST: what is lut?
    return np.array([cls.color for cls in self.classes], dtype=int)
get_class_name_from_idx 
get_class_name_from_idx(idx)

Returns the class name of the class at position idx in self.classes

Source code in niceml/data/datadescriptions/semsegdatadescritption.py 
70
71
72
def get_class_name_from_idx(self, idx: int) -> str:
    """Returns the class name of the class at position `idx` in `self.classes`"""
    return self.get_output_channel_names()[idx]
get_input_channel_count 
get_input_channel_count()

Returns the number of input channels

Source code in niceml/data/datadescriptions/semsegdatadescritption.py 
58
59
60
def get_input_channel_count(self) -> int:
    """Returns the number of input channels"""
    return self.channel_count
get_input_image_size 
get_input_image_size()

Returns the input size of the image(s)

Source code in niceml/data/datadescriptions/semsegdatadescritption.py 
50
51
52
def get_input_image_size(self) -> ImageSize:
    """Returns the input size of the image(s)"""
    return self.input_image_size
get_output_channel_names 
get_output_channel_names()

Returns the names of the output channels

Source code in niceml/data/datadescriptions/semsegdatadescritption.py 
54
55
56
def get_output_channel_names(self) -> List[str]:
    """Returns the names of the output channels"""
    return [x.name for x in self.classes]
get_output_image_size 
get_output_image_size()

Returns the output size of the image(s)

Source code in niceml/data/datadescriptions/semsegdatadescritption.py 
62
63
64
def get_output_image_size(self) -> ImageSize:
    """Returns the output size of the image(s)"""
    return self.output_image_size
get_use_void_class 
get_use_void_class()

returns bool to use background_class

Source code in niceml/data/datadescriptions/semsegdatadescritption.py 
74
75
76
def get_use_void_class(self) -> bool:
    """returns bool to use background_class"""
    return self.use_background_class

Functions

create_number_semseg_datadescription 

create_number_semseg_datadescription(max_number)

Creates a list of SemSegClassInfo for the number dataset

Source code in niceml/data/datadescriptions/semsegdatadescritption.py 
79
80
81
82
83
def create_number_semseg_datadescription(
    max_number: int,
) -> List[SemSegClassInfo]:  # QUEST: still used?
    """Creates a list of SemSegClassInfo for the number dataset"""  # QUEST: better docstring
    return [SemSegClassInfo([idx], f"{idx}") for idx in range(max_number)]