Colors

In the Dead Leaves Model, the color of each object can be sampled from different types of distributions. The colors are specified via the color_param_distribution dictionary sampling color through a LeafAppearanceSampler. Different color spaces or sources are supported, as described below.

Gray-scale

Gray-scale leaves are defined using a single channel, "gray". You can specify any supported distribution for the gray values. For example, a normal distribution centered at 0.5:

{
    "gray": <distribution>
}

• Range: 0 (black) to 1 (white), values sampled outside this range are clipped.

• Use case: Simplest and common setup for monochromatic images.

Example

Show code cell source

Hide code cell source

from deadleaves import LeafGeometryGenerator, LeafAppearanceSampler, ImageRenderer

model = LeafGeometryGenerator(
    "circular", 
    {"area": {"powerlaw": {"low": 100.0, "high": 10000.0, "k": 1.5}}},
    (256,256)
)
leaf_table, segmentation_map = model.generate_segmentation()

colormodel = LeafAppearanceSampler(leaf_table)
colormodel.sample_color({"gray": {"uniform": {"low": 0.0, "high": 1.0}}})

renderer = ImageRenderer(colormodel.leaf_table, segmentation_map)
renderer.render_image()
renderer.show(figsize = (3,3))

RGB

RGB colors are defined using three channels: "R", "G", and "B". Each channel has its own distribution, allowing for fully random or correlated color sampling:

{
    "R": <distribution>, 
    "G": <distribution>, 
    "B": <distribution>
}

• Range for each channel: 0 to 1

• Use case: Full-color images where red, green, and blue components are sampled independently or according to specific distributions.

Example

Show code cell source

Hide code cell source

from deadleaves import LeafGeometryGenerator, LeafAppearanceSampler, ImageRenderer

model = LeafGeometryGenerator(
    "circular", 
    {"area": {"powerlaw": {"low": 100.0, "high": 10000.0, "k": 1.5}}},
    (256,256)
)
leaf_table, segmentation_map = model.generate_segmentation()

colormodel = LeafAppearanceSampler(leaf_table)
colormodel.sample_color(
    {
        "R": {"uniform": {"low": 0.0, "high": 1.0}},
        "G": {"uniform": {"low": 0.0, "high": 1.0}},
        "B": {"uniform": {"low": 0.0, "high": 1.0}}
        }
)

renderer = ImageRenderer(colormodel.leaf_table, segmentation_map)
renderer.render_image()
renderer.show(figsize = (3,3))

HSV

HSV is an alternative color space that can be useful for separating hue from saturation and luminance:

{
    "H": <distribution>, 
    "S": <distribution>, 
    "V": <distribution>
}

• Range for each channel: 0 to 1

• Use case: Easily sample colors with controlled hue.

Example

Show code cell source

Hide code cell source

from deadleaves import LeafGeometryGenerator, LeafAppearanceSampler, ImageRenderer

model = LeafGeometryGenerator(
    "circular", 
    {"area": {"powerlaw": {"low": 100.0, "high": 10000.0, "k": 1.5}}},
    (256,256)
)
leaf_table, segmentation_map = model.generate_segmentation()

colormodel = LeafAppearanceSampler(leaf_table)
colormodel.sample_color(
    {
        "H": {"normal": {"loc": 0.5, "scale": 0.1}},
        "S": {"normal": {"loc": 0.5, "scale": 0.1}},
        "V": {"normal": {"loc": 0.5, "scale": 0.1}}
        }
)

renderer = ImageRenderer(colormodel.leaf_table, segmentation_map)
renderer.render_image()
renderer.show(figsize = (3,3))

From Image

You can also sample colors from an existing image. This allows the Dead Leaves to imitate real color distributions from a source:

{
    "source": {"image": {"dir": <value>}}
}

• dir: Path to the folder of images to sample from

• Use case: Generate synthetic images that match the palette of a real image.

Example

Show code cell source

Hide code cell source

from deadleaves import LeafGeometryGenerator, LeafAppearanceSampler, ImageRenderer

model = LeafGeometryGenerator(
    "circular", 
    {"area": {"powerlaw": {"low": 100.0, "high": 10000.0, "k": 1.5}}},
    (256,256)
)
leaf_table, segmentation_map = model.generate_segmentation()

colormodel = LeafAppearanceSampler(leaf_table)
colormodel.sample_color({"source": {"image": {"dir": "../../examples/images"}}})

renderer = ImageRenderer(colormodel.leaf_table, segmentation_map)
renderer.render_image()
renderer.show(figsize = (3,3))