pandora.cost_volume_confidence.ambiguity

This module contains functions for estimating confidence from ambiguity.

Classes

Ambiguity

Ambiguity class allows to estimate a confidence from the cost volume

Module Contents

class pandora.cost_volume_confidence.ambiguity.Ambiguity(**cfg: str)

Bases: pandora.cost_volume_confidence.cost_volume_confidence.AbstractCostVolumeConfidence

Ambiguity class allows to estimate a confidence from the cost volume

_ETA_MIN = 0.0

_ETA_MAX = 0.7

_ETA_STEP = 0.01

_PERCENTILE = 1.0

_NORMALIZATION = True

_method = 'ambiguity'

cfg

_eta_min

_percentile

_normalization

_eta_max

_eta_step

_indicator

_etas

_nbr_etas

check_conf(**cfg: str | float) → Dict[str, str | float]

Add default values to the dictionary if there are missing elements and check if the dictionary is correct

Parameters:

cfg (dict) – ambiguity configuration

Return cfg:

ambiguity configuration updated

Return type:

dict

desc() → None

Describes the confidence method :return: None

confidence_prediction(disp: xarray.Dataset, img_left: xarray.Dataset = None, img_right: xarray.Dataset = None, cv: xarray.Dataset = None) → Tuple[xarray.Dataset, xarray.Dataset]

Computes a confidence measure that evaluates the matching cost function at each point

Parameters:

• disp (xarray.Dataset) – the disparity map dataset

• img_left – left Dataset image

• img_right (xarray.Dataset) – right Dataset image

• cv (xarray.Dataset) – cost volume dataset

Tye img_left:

xarray.Dataset

Returns:

the disparity map and the cost volume with a new indicator ‘ambiguity_confidence’ in the DataArray confidence_measure

Return type:

Tuple(xarray.Dataset, xarray.Dataset) with the data variables:

• confidence_measure 3D xarray.DataArray (row, col, indicator)

normalize_with_percentile(ambiguity: numpy.ndarray) → numpy.ndarray

Normalize ambiguity with percentile

Parameters:

ambiguity (2D np.ndarray (row, col) dtype = float32) – ambiguity

Returns:

the normalized ambiguity

Return type:

2D np.ndarray (row, col) dtype = float32

static compute_ambiguity(cv: numpy.ndarray, etas: numpy.ndarray, nbr_etas: int, grids: numpy.ndarray, disparity_range: numpy.ndarray, type_measure_min: bool) → numpy.ndarray

Computes ambiguity.

Parameters:

• cv (3D np.ndarray (row, col, disp)) – cost volume

• etas (np.ndarray) – range between eta_min and eta_max with step eta_step

• nbr_etas (int) – number of etas

• grids (2D np.ndarray (min, max)) – array containing min and max disparity grids

• disparity_range (np.ndarray) – array containing disparity range

• type_measure_min (bool) – True for min and False for max

Returns:

the normalized ambiguity

Return type:

2D np.ndarray (row, col) dtype = float32

static compute_ambiguity_and_sampled_ambiguity(cv: numpy.ndarray, etas: numpy.ndarray, nbr_etas: int, grids: numpy.ndarray, disparity_range: numpy.ndarray)

Return the ambiguity and sampled ambiguity, useful for evaluating ambiguity in notebooks

Parameters:

• cv (3D np.ndarray (row, col, disp)) – cost volume

• etas (np.ndarray) – range between eta_min and eta_max with step eta_step

• nbr_etas (int) – nuber of etas

• grids (2D np.ndarray (min, max)) – array containing min and max disparity grids

• disparity_range (np.ndarray) – array containing disparity range

Returns:

the normalized ambiguity and sampled ambiguity

Return type:

Tuple(2D np.ndarray (row, col) dtype = float32, 3D np.ndarray (row, col) dtype = float32)