View source on GitHub

Adois

Tiles processor that uses the adois model to detect and classify impervious surfaces.

adois is a segmentation model that is trained to detect and classify impervious surfaces from digital orthophotos. It differentiates between non-impervious surfaces (e.g., vegetation, soil or water), buildings and impervious surfaces (e.g., pavements, roads, sidewalks, driveways, parking lots or industrial areas). It is recommended to use the model with leaf-off orthophotos (i.e., without foliage on trees or shrubs), so canopies do not cover buildings or impervious surfaces.

Notes

• Accessing a channel by its name assumes the time step is None

Model input channels

• ChannelName.R: Red channel, raster channel, ground sampling distance of 0.2 meters, normalized values in range [0, 1]
• ChannelName.G: Green channel, raster channel, ground sampling distance of 0.2 meters, normalized values in range [0, 1]
• ChannelName.B: Blue channel, raster channel, ground sampling distance of 0.2 meters, normalized values in range [0, 1]
• ChannelName.NIR: Near-infrared channel, raster channel, ground sampling distance of 0.2 meters, normalized values in range [0, 1]

Model output channels

• 'adois': Impervious surfaces channel, raster channel, ground sampling distance of 0.2 meters, the classes are non-impervious surfaces (value 0), buildings (value 1), and impervious surfaces (value 2)

Additional dependencies

adois requires huggingface_hub and onnxruntime as additional dependencies.

Implements the TilesProcessor protocol.

PARAMETER	DESCRIPTION
r_channel_key	Channel name or channel name and time step combination of the red channel TYPE: ChannelName | str | ChannelKey DEFAULT: ChannelName.R
g_channel_key	Channel name or channel name and time step combination of the green channel TYPE: ChannelName | str | ChannelKey DEFAULT: ChannelName.G
b_channel_key	Channel name or channel name and time step combination of the blue channel TYPE: ChannelName | str | ChannelKey DEFAULT: ChannelName.B
nir_channel_key	Channel name or channel name and time step combination of the near-infrared channel TYPE: ChannelName | str | ChannelKey DEFAULT: ChannelName.NIR
out_channel_key	Channel name or channel name and time step combination of the output channel TYPE: ChannelName | str | ChannelKey DEFAULT: 'adois'
cache_dir_path	Path to the cache directory of the model TYPE: Path DEFAULT: Path('cache')
remove_channels	If True, the channels are removed TYPE: bool DEFAULT: True

from_config classmethod

Creates the adois model from the configuration.

PARAMETER	DESCRIPTION
config	Configuration TYPE: AdoisConfig

RETURNS	DESCRIPTION
Adois	Adois

__call__

Runs the adois model.

PARAMETER	DESCRIPTION
tiles	Tiles TYPE: Tiles

RETURNS	DESCRIPTION
Tiles	Tiles

---

AdoisConfig

Bases: pydantic.BaseModel

Configuration for the from_config class method of Adois

Create the configuration from a config file

• Use null instead of None
• Use false or true instead of False or True

Example

You can create the configuration from a config file.

config.yaml

package: 'aviary'
name: 'Adois'
config:
  r_channel_key: 'r'
  g_channel_key: 'g'
  b_channel_key: 'b'
  nir_channel_key: 'nir'
  out_channel_key: 'adois'
  cache_dir_path: 'cache'
  remove_channels: true

ATTRIBUTE	DESCRIPTION
r_channel_key	Channel name or channel name and time step combination of the red channel - defaults to ChannelName.R TYPE: ChannelName | str | ChannelKey
g_channel_key	Channel name or channel name and time step combination of the green channel - defaults to ChannelName.G TYPE: ChannelName | str | ChannelKey
b_channel_key	Channel name or channel name and time step combination of the blue channel - defaults to ChannelName.B TYPE: ChannelName | str | ChannelKey
nir_channel_key	Channel name or channel name and time step combination of the near-infrared channel - defaults to ChannelName.NIR TYPE: ChannelName | str | ChannelKey
out_channel_key	Channel name or channel name and time step combination of the output channel - defaults to 'adois' TYPE: ChannelName | str | ChannelKey
cache_dir_path	Path to the cache directory of the model - defaults to 'cache' TYPE: Path
remove_channels	If True, the channels are removed - defaults to True TYPE: bool