Colormap#

Render a single-channel image with a color map, e.g. to visualize depth, heat or edge maps.

Functions#

kornia.color.apply_colormap(input_tensor, colormap)[source]#

Color each channel of an image with a colormap.

Try in browser

out = apply_colormap(input_tensor, colormap) replaces each value of each channel with a color of colormap.

Convention:: Accepts a rank-3 (C, H, W) or rank-4 (B, C, H, W) tensor and returns (B, 3*C, H, W), with B = 1 for rank-3 input; input channel c becomes output channels 3c to 3c + 2. Integer inputs use [0, 255]. Floating inputs use [0, 1] or [0, 255] independently for each sample and channel, based on that channel’s maximum. A value x in [0, 1] selects entry floor(x * N) of the N palette colors, and x = 1 selects the last, as in matplotlib, so every color covers an equal share of the range. Values below or above the range select the first or the last color.

Parameters:

input_tensor (Tensor) – image with shape (C, H, W) or (B, C, H, W).
colormap (ColorMap) – palette to apply.

Return type:

Tensor

Returns:

Color image with shape (B, 3 * C, H, W) and the dtype of the palette.

See also

ApplyColorMap: the module form.
ColorMap: builds the palette.

Example

>>> input_tensor = torch.tensor([[[0, 1, 2], [15, 25, 33], [128, 158, 188]]])
>>> colormap = ColorMap(base=ColorMapType.autumn)
>>> apply_colormap(input_tensor, colormap)
tensor([[[[1.0000, 1.0000, 1.0000],
          [1.0000, 1.0000, 1.0000],
          [1.0000, 1.0000, 1.0000]],

         [[0.0000, 0.0000, 0.0000],
          [0.0476, 0.0952, 0.1270],
          [0.5079, 0.6190, 0.7460]],

         [[0.0000, 0.0000, 0.0000],
          [0.0000, 0.0000, 0.0000],
          [0.0000, 0.0000, 0.0000]]]])

Modules#

class kornia.color.ApplyColorMap(colormap)[source]#

Color each channel of an image with a colormap (module form).

Try in browser

cmap = ApplyColorMap(colormap) stores the palette as the buffer colors, and out = cmap(input_tensor) is apply_colormap() with it.

Convention:

See the convention block of apply_colormap().

Parameters:: colormap (ColorMap) – palette to apply.

Shape:

input_tensor: (C, H, W) or (B, C, H, W)
output: (B, 3 * C, H, W)

See also

apply_colormap(): the one-call form.
ColorMap: builds the palette.

Example

>>> input_tensor = torch.tensor([[[0, 1, 2], [15, 25, 33], [128, 158, 188]]])
>>> colormap = ColorMap(base=ColorMapType.autumn)
>>> ApplyColorMap(colormap=colormap)(input_tensor)
tensor([[[[1.0000, 1.0000, 1.0000],
          [1.0000, 1.0000, 1.0000],
          [1.0000, 1.0000, 1.0000]],

         [[0.0000, 0.0000, 0.0000],
          [0.0476, 0.0952, 0.1270],
          [0.5079, 0.6190, 0.7460]],

         [[0.0000, 0.0000, 0.0000],
          [0.0000, 0.0000, 0.0000],
          [0.0000, 0.0000, 0.0000]]]])

class kornia.color.ColorMap(base, num_colors=64, device=None, dtype=None)[source]#

Color palette for apply_colormap().

cmap = ColorMap(base, num_colors) builds a built-in colormap, or one from your own RGB triples, with num_colors entries and stores it as cmap.colors.

Convention:: colors has shape (3, num_colors) in RGB channel order. Built-in names are case-insensitive; a custom base is an ordered list of RGB triples in \([0, 1]\) and is linearly resampled to num_colors. For num_colors > 1, entry i sits at i / (num_colors - 1) along the base, so the first and last colors of the base are the first and last of the map. With num_colors == 1, the map holds the base’s first color.

Parameters:

base (Union[list[List[float]], str, ColorMapType]) – list of RGB triples, or a built-in colormap as a ColorMapType or its name.
num_colors (int, optional) – number of palette entries. Default: 64
device (Optional[device], optional) – device of the palette. Default: None
dtype (Optional[dtype], optional) – dtype of the palette. Defaults to torch.get_default_dtype(). Default: None

See also

ColorMapType: the built-in colormaps.
apply_colormap(): colors an image with the palette.
ApplyColorMap: the module form of apply_colormap().

Examples

>>> ColorMap(base='viridis', num_colors=8).colors
tensor([[0.2670, 0.2752, 0.2124, 0.1534, 0.1223, 0.2889, 0.6266, 0.9932],
        [0.0049, 0.1949, 0.3597, 0.4970, 0.6332, 0.7584, 0.8546, 0.9062],
        [0.3294, 0.4960, 0.5517, 0.5577, 0.5304, 0.4284, 0.2234, 0.1439]])

>>> ColorMap(base=[[0., 0.5 , 1.0], [1., 0.5, 0.]], num_colors=8).colors
tensor([[0.0000, 0.1429, 0.2857, 0.4286, 0.5714, 0.7143, 0.8571, 1.0000],
        [0.5000, 0.5000, 0.5000, 0.5000, 0.5000, 0.5000, 0.5000, 0.5000],
        [1.0000, 0.8571, 0.7143, 0.5714, 0.4286, 0.2857, 0.1429, 0.0000]])

kornia.color.RGBColor#: alias of List[float]

Color maps available:

class kornia.color.ColorMapType(value)[source]#

Names of the built-in colormaps.

ColorMapType.viridis, or its name 'viridis', selects a built-in colormap in ColorMap. list() returns all names.

See also

ColorMap: builds the palette of a member.

autumn = 1#

bone = 2#

jet = 3#

winter = 4#

rainbow = 5#

ocean = 6#

summer = 7#

spring = 8#

cool = 9#

hsv = 10#

brg = 11#

pink = 12#

hot = 13#

plasma = 14#

viridis = 15#

cividis = 16#

twilight = 17#

turbo = 18#

seismic = 19#

classmethod list()[source]#

Return a list of names of enumeration members.

Return type:: list[str]
Returns:: A list containing the names of enumeration members.