Subsampling

The subsampling pipeline takes sensor images as inputs and generates subsampled versions at the resolutions specified in the configuration. A sensor image may be accompanied by classifications, which will also be downsampled. The list of resolutions this pipeline outputs is the list that will be used in the meta pipeline.

This pipeline does not run at multiple resolutions.

Allowed inputs

Subsampling can take only sensor images as inputs (and their respective classification), as shown in the input section of the documentation.

Applications

Name: “sensors_subsampling”

Description

Subsample the sensor images.

Configuration

Name	Description	Type	Available value	Default value	Required
method	Method for sensors_subsampling	string	“rasterio”	“rasterio”	No
tile_size	Size of tiling	int		10000	No
interpolator_image	Interpolator that should be used for the image	str		“bilinear”	No
interpolator_classif	Interpolator that should be used for the classif	str		“nearest”	No
interpolator_mask	Interpolator that should be used for the mask	str		“nearest”	No

Advanced Parameters

Name	Description	Type	Default value
save_intermediate_data	Save intermediate data for all applications inside this pipeline.	bool	False
resolutions	The resolution(s) at which the outputs should be resampled.	list[int]	[16, 4, 1]
min_image_size	Image size below which subsampling is not done.	int	80

Below is an example configuration for this pipeline :

YAML JSON Copy

subsampling:
  applications:
    sensors_subsampling:
      method: "rasterio"

  advanced:
    save_intermediate_data: true
    resolutions: [32, 8, 1]

{
    "subsampling": {
        "applications": {
            "sensors_subsampling": {
                "method": "rasterio"
            }
        },
        "advanced": {
            "save_intermediate_data": true,
            "resolutions": [
                32,
                8,
                1
            ]
        }
    }
}