Model

`Model`

Bases: Protocol

Protocol for models

Models are callables that transform inputs into predictions.

Currently implemented models

ONNXSegmentationModel: ONNX model for segmentation

`call`

Runs the model.

PARAMETER	DESCRIPTION
`inputs`	batched inputs TYPE: `npt.NDArray`

RETURNS	DESCRIPTION
`npt.NDArray`	batched predictions

`SegmentationModel`

Model for segmentation

PARAMETER	DESCRIPTION
`path`	path to the model TYPE: `Path`
`ground_sampling_distance`	ground sampling distance in meters TYPE: `GroundSamplingDistance`
`buffer_size`	buffer size in meters (specifies the area around the tile that is additionally fetched) TYPE: `BufferSize` DEFAULT: `0`
`device`	device (`CPU` or `CUDA`) TYPE: `Device` DEFAULT: `Device.CPU`

`from_huggingface` `classmethod`

Creates a segmentation model from the Hugging Face Hub.

PARAMETER	DESCRIPTION
`repo`	repository (Hugging Face Hub, e.g., 'user/repo') TYPE: `str`
`path`	path to the model (Hugging Face Hub) TYPE: `str`
`ground_sampling_distance`	ground sampling distance in meters TYPE: `GroundSamplingDistance`
`buffer_size`	buffer size in meters (specifies the area around the tile that is additionally fetched) TYPE: `BufferSize` DEFAULT: `0`
`device`	device (`CPU` or `CUDA`) TYPE: `Device` DEFAULT: `Device.CPU`

RETURNS	DESCRIPTION
`SegmentationModel`	segmentation model

`from_aviary` `classmethod`

Creates a segmentation model from the name of a model in aviary.

PARAMETER	DESCRIPTION
`name`	name of the model TYPE: `str`
`ground_sampling_distance`	ground sampling distance in meters TYPE: `GroundSamplingDistance`
`buffer_size`	buffer size in meters (specifies the area around the tile that is additionally fetched) TYPE: `BufferSize` DEFAULT: `0`
`device`	device (`CPU` or `CUDA`) TYPE: `Device` DEFAULT: `Device.CPU`

RETURNS	DESCRIPTION
`SegmentationModel`	segmentation model

`from_config` `classmethod`

Creates a segmentation model from the configuration.

PARAMETER	DESCRIPTION
`config`	configuration TYPE: `SegmentationModelConfig`

RETURNS	DESCRIPTION
`SegmentationModel`	segmentation model

RAISES	DESCRIPTION
`AviaryUserError`	Invalid configuration

`call`

Runs the model.

PARAMETER	DESCRIPTION
`inputs`	batched inputs TYPE: `npt.NDArray`

RETURNS	DESCRIPTION
`npt.NDArray`	batched predictions

`SegmentationModelConfig`

Bases: pydantic.BaseModel

Configuration for the from_config class method of SegmentationModel

The configuration must have one of the following field sets

name
repo and path
path

ATTRIBUTE	DESCRIPTION
`path`	path to the model (local or Hugging Face Hub) TYPE: `str \| None`
`repo`	repository (Hugging Face Hub, e.g., 'user/repo') TYPE: `str \| None`
`name`	name of the model TYPE: `str \| None`
`ground_sampling_distance`	ground sampling distance in meters TYPE: `GroundSamplingDistance`
`buffer_size`	buffer size in meters (specifies the area around the tile that is additionally fetched) TYPE: `BufferSize`
`device`	device ('cpu' or 'cuda') TYPE: `Device`

`ONNXSegmentationModel`

Bases: SegmentationModel

ONNX model for segmentation

Implements the Model protocol.

PARAMETER	DESCRIPTION
`path`	path to the model TYPE: `Path`
`ground_sampling_distance`	ground sampling distance in meters TYPE: `GroundSamplingDistance`
`buffer_size`	buffer size in meters (specifies the area around the tile that is additionally fetched) TYPE: `BufferSize` DEFAULT: `0`
`device`	device (`CPU` or `CUDA`) TYPE: `Device` DEFAULT: `Device.CPU`

`from_huggingface` `classmethod`

Creates a segmentation model from the Hugging Face Hub.

PARAMETER	DESCRIPTION
`repo`	repository (Hugging Face Hub, e.g., 'user/repo') TYPE: `str`
`path`	path to the model (Hugging Face Hub) TYPE: `str`
`ground_sampling_distance`	ground sampling distance in meters TYPE: `GroundSamplingDistance`
`buffer_size`	buffer size in meters (specifies the area around the tile that is additionally fetched) TYPE: `BufferSize` DEFAULT: `0`
`device`	device (`CPU` or `CUDA`) TYPE: `Device` DEFAULT: `Device.CPU`

RETURNS	DESCRIPTION
`SegmentationModel`	segmentation model

`from_aviary` `classmethod`

Creates a segmentation model from the name of a model in aviary.

PARAMETER	DESCRIPTION
`name`	name of the model TYPE: `str`
`ground_sampling_distance`	ground sampling distance in meters TYPE: `GroundSamplingDistance`
`buffer_size`	buffer size in meters (specifies the area around the tile that is additionally fetched) TYPE: `BufferSize` DEFAULT: `0`
`device`	device (`CPU` or `CUDA`) TYPE: `Device` DEFAULT: `Device.CPU`

RETURNS	DESCRIPTION
`SegmentationModel`	segmentation model

`from_config` `classmethod`

Creates a segmentation model from the configuration.

PARAMETER	DESCRIPTION
`config`	configuration TYPE: `SegmentationModelConfig`

RETURNS	DESCRIPTION
`SegmentationModel`	segmentation model

RAISES	DESCRIPTION
`AviaryUserError`	Invalid configuration

`call`

Runs the model.

PARAMETER	DESCRIPTION
`inputs`	batched inputs TYPE: `npt.NDArray`

RETURNS	DESCRIPTION
`npt.NDArray`	batched predictions

Model

__call__

SegmentationModel

from_huggingface classmethod

from_aviary classmethod

from_config classmethod

__call__

SegmentationModelConfig

ONNXSegmentationModel

from_huggingface classmethod

from_aviary classmethod

from_config classmethod

__call__

`Model`

`call`

`SegmentationModel`

`from_huggingface` `classmethod`

`from_aviary` `classmethod`

`from_config` `classmethod`

`call`

`SegmentationModelConfig`

`ONNXSegmentationModel`

`from_huggingface` `classmethod`

`from_aviary` `classmethod`

`from_config` `classmethod`

`call`