SLURP Configuration

SLURP can be configured with JSON files that allows you to set various parameters.

The main configuration file is given to the SLURP’s API or CLI through the main_config argument.

Tip

Some arguments defined by the JSON file can be overwritten by command line arguments or by API’s arguments. For more information, please refer to API’s documentation or refer to the help of your CLI command (cli_command_name –help).

Below is an example of a configuration file:

Field Group	Field Name	Expected Type	Description	Status
input			Input data for the configuration	Mandatory
	file_vhr	Str	Input 4 bands VHR (Very High Resolution) image	Mandatory
	sensor_mode	Bool	True if input image is in its raw (sensor) geometry, False if input image is georeferenced (orthorectification)	Mandatory
aux_layers			Auxiliary layers used in processing	Mandatory
	valid_stack	Str (Default None)	Path to store the valid stack file	Optional
	file_ndvi	Str	Path to the NDVI layer	Mandatory
	file_ndwi	Str	Path to the NDWI layer	Mandatory
	extracted_pekel	Str	Path to the extracted Pekel data	Mandatory
	extracted_hand	Str	Path to the extracted HAND data	Mandatory
	extracted_wsf	Str	Path to the extracted World Settlement Footprint	Mandatory
	extracted_wbm	Str (Default None)	Path to the extracted Water Body Mask	Optional
	file_texture	Str	Path to store the texture file	Mandatory
	mnh	Str (Default None)	Path to the Mean Height of Nearest Neighbors (MN) layer	Optional
	file_cloud_gml	Str (Default None)	GML file containing cloud masks (optional)	Optional
masks			Masks for different types of land cover and features	Mandatory
	watermask	Str	For watermask computation: Output classification filename. For stackmask: Water mask to stack. Otherwise, if given, output mask will exclude water areas	Mandatory
	urbanmask	Str	For urbanmask computation: Output classification filename. For stackmask: Vegetation mask to stack.	Mandatory
	vegetationmask	Str	For vegetationmask computation : Output classification filename. For stackmask: Vegetation mask to stack. Otherwise, if given, output mask will exclude vegetation areas	Mandatory
	shadowmask	Str	For shadowmask computation : Output classification filename. For stackmask: Shadow mask to stack. Otherwise, if given, big shadow areas will be marked as background	Mandatory
	stackmask	Str	Output classification filename	Mandatory
resources			Resources configuration for parallel processing	Mandatory
	debug	Bool (Default False)	Flag for debugging mode (default : False)	Optional
	n_workers	Int	Number of CPU	Mandatory
	tile_max_size	Int	Maximum size of tiles to process	Mandatory
	multiproc_context	Str (Default ‘spawn’)	Multiprocessing strategy: ‘fork’ or ‘spawn’ for EOScale	Optional
	n_jobs	Int	Nb of parallel jobs for Random Forest (1 is recommanded : use n_workers to optimize parallel computing)	Mandatory
	save_mode	Str	Save all files (debug) or only output mask (none)	Mandatory
prepare			Preparation settings for preprocessing	Mandatory
	red	Int	Red band index	Mandatory
	green	Int	Green band index	Mandatory
	nir	Int	NIR band index	Mandatory
	cloud_mask	Str (Default None)	Path to the input cloud mask	Optional
	pekel_method	Str	Method for Pekel recovery : ‘all’ for global file and ‘month’ for monthly recovery	Mandatory
	pekel	Str	Path of the global Pekel (Global Surface Water) file	Mandatory
	pekel_monthly_occurrence	Str (Default None)	Path of the root of monthly occurrence Pekel files	Optional
	pekel_obs	Str (Default None)	Month of the desired Pekel (Global Surface Water) file (pekel_method = month)	Optional
	hand	Str	Path of the global Height Above Nearest Drainage (HAND) file	Mandatory
	wsf	Str	Path of the global World Settlement Footprint (WSF) file	Mandatory
	wbm	Str	Path of the global Water Body Mask (WBM) file	Mandatory
	texture_rad	Int	Radius for texture (std convolution) computation	Mandatory
	dtm	Str (Default None)	Digital Terrain Model, used only in sensor mode	Optional
	geoid_file	Str	Geoid file, used only in sensor mode	Mandatory
	analyse_glcm	Bool	Use a global land cover map to calculate the better number of vegetation cluster to use for mask computation	Mandatory
	land_cover_map	Str	Input land cover map, only used if ‘analyse_glcm’ is True	Mandatory
	cropped_land_cover_map	Bool	If the land_cover_map image is cropped to the input VHR file or not	Mandatory
	effective_used_config	Str	Path to the effective configuration used	Mandatory
post_process			Settings for post-processing	Mandatory
	binary_opening	Int	Size of disk structuring element	Mandatory
	binary_closing	Int	Size of disk structuring element	Mandatory
	binary_dilation	Int	Size of disk structuring element	Mandatory
	remove_small_objects	Int	The maximum area, in pixels, of a contiguous object that will be removed	Mandatory
	remove_small_holes	Int	The maximum area, in pixels, of a contiguous hole that will be filled	Mandatory
	area_closing	Str (Default None)	Area closing removes all dark structures	Optional
shadows			Shadow detection settings	Mandatory
	th_rgb	Float	Relative shadow threshold for RGB bands	Mandatory
	th_nir	Float	Relative shadow threshold for NIR band	Mandatory
	percentile	Int	Percentile value to cut histogram and estimate shadow threshold	Mandatory
	absolute_threshold	Bool	Compute shadow mask with a unique absolute threshold	Mandatory
urban			Urban classification settings	Mandatory
	files_layers	List[Str] (Default None)	Add layers as additional features used by learning algorithm	Optional
	vegmask_min_value	Int (Default None)	Vegetation min value for vegetated areas : all pixels with lower value will be predicted	Optional
	veg_binary_dilation	Int	Size of disk structuring element (dilate non vegetated areas)	Mandatory
	value_classif	Int	Input ground truth class to consider in the input ground truth	Mandatory
	gt_binary_erosion	Int	Size of disk structuring element (erode GT before picking-up samples)	Mandatory
	nb_samples_other	Int	Number of samples in other for learning	Mandatory
	nb_samples_urban	Int	Number of samples in buildings for learning	Mandatory
	max_depth	Int	Maximum depth of the decision tree	Mandatory
	nb_estimators	Int	Number of trees in Random Forest	Mandatory
vegetation			Vegetation classification settings	Mandatory
	texture_mode	Str	Labelize vegetation with (yes) or without (no) distinction low/high, or get all vegetation clusters without distinction low/high (debug)	Mandatory
	filter_texture	Int	Percentile for texture (between 1 and 99)	Mandatory
	slic_seg_size	Int	Approximative segment size	Mandatory
	slic_compactness	Float	Balance between color and space proximity (see skimage.slic documentation)	Mandatory
	nb_clusters_veg	Int	Nb of clusters considered as vegetation (1-NB_CLUSTERS)	Mandatory
	min_ndvi_veg	Float (Default None)	Minimal mean NDVI value to consider a cluster as vegetation (overload nb clusters choice)	Optional
	max_ndvi_noveg	Float (Default None)	Maximal mean NDVI value to consider a cluster as non-vegetation (overload nb clusters choice)	Optional
	non_veg_clusters	List[Int] (Default None)	Labelize each ‘non vegetation cluster’ as 0, 1, 2 (..) instead of single label (0)	Optional
	nb_clusters_low_veg	Int	Nb of clusters considered as low vegetation (1-NB_CLUSTERS)	Mandatory
	max_texture_th	Float (Default None)	Maximal texture value to consider a cluster as low vegetation (overload nb clusters choice)	Optional
water			Water detection settings	Mandatory
	files_layers	List[Str]	Add layers as additional features used by learning algorithm	Mandatory
	thresh_pekel	Int	Threshold for Pekel water occurrence detection	Mandatory
	thresh_hand	Int	Hand Threshold int >= 0	Mandatory
	hand_strict	Bool	Use not(pekelxx) for other (no water) samples	Mandatory
	strict_thresh	Int	Pekel Threshold float if hand_strict	Mandatory
	simple_ndwi_threshold	Bool	Compute water mask as a simple NDWI threshold - useful in arid places where no water is known by Peckel	Mandatory
	ndwi_threshold	Float	Threshold used when Pekel is empty in the area	Mandatory
	samples_method	Str	Select method for choosing learning samples (‘grid’ or ‘random’ or ‘smart’)	Mandatory
	nb_samples_water	Int	Number of samples in water for learning	Mandatory
	nb_samples_other	Int	Number of samples in other for learning	Mandatory
	nb_samples_auto	Bool	Auto select number of samples for water and other	Mandatory
	auto_pct	Float	Percentage of samples points, to use with -nb_samples_auto	Mandatory
	smart_area_pct	Int	For smart method, importance of area for selecting number of samples in each water surface	Mandatory
	smart_minimum	Int	For smart method, minimum number of samples in each water surface.	Mandatory
	grid_spacing	Int	For grid method, select samples on a regular grid (40 pixels seems to be a good value)	Mandatory
	max_depth	Int	Max depth of trees	Mandatory
	nb_estimators	Int	Number of estimators for the water classifier	Mandatory
	no_pekel_filter	Bool	Deactivate postprocess with pekel which only keeps surfaces already known by pekel	Mandatory
	hand_filter	Bool	Postprocess with Hand (set to 0 when hand > thresh), incompatible with hand_strict	Mandatory
	value_classif	Int	Output classification value (default is 1)	Mandatory
stack			Stack processing settings	Mandatory
	building_threshold	Int	Threshold for building detection in the stack	Mandatory
	building_erosion	Int	Supposed buildings will be eroded by this size in the marker step	Mandatory
	bonus_gt	Int	Bonus for pixels covered by GT, in the watershed regularization step (ex : +30 to improve discrimination between building and background)	Mandatory
	malus_shadow	Int	Value of the malus for pixels in shadow, in the watershed regularization step	Mandatory
	value_classif_low_veg	Int (Default 1)	Output classification value for low vegetation	Optional
	value_classif_high_veg	Int (Default 2)	Output classification value for high vegetation	Optional
	value_classif_water	Int (Default 3)	Output classification value for water	Optional
	value_classif_buildings	Int (Default 4)	Output classification value for buildings	Optional
	value_classif_bare_ground	Int (Default 6)	Output classification value for bare ground	Optional
	value_classif_sea	Int (Default 7)	Output classification value for sea	Optional
	value_classif_lake	Int (Default 8)	Output classification value for sea	Optional
	value_classif_river	Int (Default 9)	Output classification value for sea	Optional
	value_classif_false_positive_buildings	Int	Output classification value for buildings false positive	Mandatory
	value_classif_background	Int	Output classification value for background	Mandatory
	vegmask_min_value	Int	Maximum allowed value in the vegetation mask	Mandatory
	binary_closing	Int	Size of disk structuring element	Mandatory
	binary_opening	Int	Size of disk structuring element	Mandatory
	categorized_watermask	Bool (Default False)	If true, stack_mask will infer water body category (lake, river, sea, unknown) from a general water body mask	Optional
	minimal_size_water_area	Int (Default 10000)	Minimal area (in pixels) of water bodies	Optional