HeavyEdge-Features documentation#

HeavyEdge-Features is a Python package for quantifying shape features from coating edge profiles.

Usage#

HeavyEdge-Features can be used either as a command line program or as a Python module.

Command line#

Command line interface provides pre-defined subroutines for training and prediction. It can be invoked by:

heavyedge features-global <args>
heavyedge features-local <args>

Refer to help message of each command for their arguments.

Python module#

The Python module heavyedge_features provides functions and classes for Python runtime. Refer to Runtime API section for high-level interface.

Module reference#

This section provides reference for heavyedge_features Python module.

Runtime API#

High-level Python runtime interface.

heavyedge_features.api.global_deviation(soft_labels, target_indices, logger=<function <lambda>>)[source]#

Compute global shape deviations using probabilistic classification labels.

Negative values indicaete profiles are within the desired classes. Larger values mean more deviation from the desired classes.

Parameters:

soft_labelsnp.ndarray: Probabilistic classification labels for the profiles.
target_indiceslist of int: Indices of target classes to compute values for.
loggercallable, optional: Logger function which accepts a progress message string.

Returns:

valuesnp.ndarray: Array containing global shape deviations for each profile.

Examples

>>> import numpy as np
>>> from heavyedge_classify.samples import get_sample_path
>>> from heavyedge_features.api import global_deviation
>>> soft_labels = np.load(get_sample_path("labels-pred.npy"))
>>> global_deviation(soft_labels, [0]).shape
(75,)

heavyedge_features.api.edge_height(profiles, logger=<function <lambda>>)[source]#

Dimensionless edge height of edge profiles.

Parameters:

profilesheavyedge.ProfileData: Open h5 file of profiles.
loggercallable, optional: Logger function which accepts a progress message string.

Returns:

heightsnp.ndarray: Array containing edge height values for each profile.

Examples

>>> from heavyedge import ProfileData
>>> from heavyedge_features.samples import get_sample_path as features_sample
>>> from heavyedge_features.api import edge_height
>>> edge_height(ProfileData(features_sample("Profiles.h5"))).shape
(75,)

heavyedge_features.api.edge_width(profiles, hard_labels, wet_thicknesses, sigma, type1_indices, type2_indices, type3_indices, logger=<function <lambda>>)[source]#

Detect edge with of profiles using profile data and classification labels.

Parameters:

profilesheavyedge.ProfileData: Open h5 file of profiles.
hard_labelsnp.ndarray: Hard classification labels for the profiles.
wet_thicknessesnp.ndarray: Wet thickness values for the profiles.
sigmascalar: Standard deviation of Gaussian filter for smoothing. Using the same value as the one used for preprocessing is recommended.
type1_indices, type2_indices, type3_indiceslist of int: Lists of indices of Type 1, 2, and 3 classes from trained labels, respectively.
loggercallable, optional: Logger function which accepts a progress message string.

Returns:

widthsnp.ndarray: Array containing edge width values for each profile.

Examples

>>> from heavyedge import ProfileData
>>> from heavyedge_classify.samples import get_sample_path as classify_sample
>>> from heavyedge_features.samples import get_sample_path as features_sample
>>> from heavyedge_features.api import edge_width
>>> import numpy as np
>>> profiles = ProfileData(features_sample("Profiles.h5"))
>>> hard_labels = np.load(classify_sample("labels-pred.npy")).argmax(axis=1)
>>> wet_thicknesses = np.full(hard_labels.shape, 0.25)
>>> sigma = 32
>>> edge_width(profiles, hard_labels, wet_thicknesses, sigma, [0], [1], [2]).shape
(75,)

Low-level API#

Information projection to allowed classes.

heavyedge_features.iproj.signed_iproj(p, target_indices)[source]#

Signed information projection distance to allowed classes.

Parameters:

parray-like, shape (K,): Probability distribution over K classes.
target_indiceslist of int: List of target class indices to project onto.

Returns:

signed_distancefloat: Signed information projection distance to the allowed classes.
projected_distributionarray-like, shape (K,): Probability distribution after projection.

Examples

>>> import numpy as np
>>> from heavyedge_features.iproj import signed_iproj
>>> p = np.array([0.1, 0.7, 0.2])
>>> target_indices = [0, 2]
>>> dist, q = signed_iproj(p, target_indices)
>>> dist
 np.float64(0.164...)
>>> q
array([0.117..., 0.441..., 0.441...])

Edge width by edge type.

heavyedge_features.edge_width.width_type0(x, Y, L, wt)[source]#

Edge width for type 0 profiles.

Parameters:

xarray of shape (M,): X grid of profiles.
Yarray of shape (M,): Height data of profile.
Lint: Length of profile before contact point in number of points.
wtscalar: Wet thickness of the profile.

Returns:

scalar: Edge width of the profile.

heavyedge_features.edge_width.width_type1(x, Y, L, wt)[source]#

Edge width for type 1 profiles.

Parameters:

xarray of shape (M,): X grid of profiles.
Yarray of shape (M,): Height data of profile.
Lint: Length of profile before contact point in number of points.
wtscalar: Wet thickness of the profile.

Returns:

scalar: Edge width of the profile.

Examples

>>> from heavyedge import get_sample_path, ProfileData
>>> from heavyedge_features.edge_width import width_type1
>>> with ProfileData(get_sample_path("Prep-Type1.h5")) as data:
...     x = data.x()
...     Y, L, _ = data[0]
>>> wt = 0.25
>>> b = width_type1(x, Y, L, wt)
>>> import matplotlib.pyplot as plt
... plt.plot(x, Y, color="gray", alpha=0.5)
... plt.axhline(wt, ls="--", label="Wet thickness")
... plt.axvline(x[L - 1] - b, color="red", ls="--", label="Edge boundary")
... plt.legend()

heavyedge_features.edge_width.width_type2(x, Y, L, sigma)[source]#

Edge width for type 2 profiles.

Parameters:

xarray of shape (M,): X grid of profiles.
Yarray of shape (M,): Height data of profile.
Lint: Length of profile before contact point in number of points.
sigmascalar: Standard deviation of Gaussian filter for smoothing.

Returns:

scalar: Edge width of the profile.

Examples

>>> from heavyedge import get_sample_path, ProfileData
>>> from heavyedge_features.edge_width import width_type2
>>> with ProfileData(get_sample_path("Prep-Type2.h5")) as data:
...     x = data.x()
...     Y, L, _ = data[0]
>>> sigma = 32.0
>>> b = width_type2(x, Y, L, sigma)
>>> import matplotlib.pyplot as plt
... plt.plot(x, Y, color="gray", alpha=0.5)
... plt.axvline(x[L - 1] - b, color="red", ls="--", label="Edge boundary")
... plt.legend()

heavyedge_features.edge_width.width_type3(x, Y, L, sigma)[source]#

Edge width for type 3 profiles.

Parameters:

xarray of shape (M,): X grid of profiles.
Yarray of shape (M,): Height data of profile.
Lint: Length of profile before contact point in number of points.
sigmascalar: Standard deviation of Gaussian filter for smoothing.

Returns:

scalar: Edge width of the profile.

Examples

>>> from heavyedge import get_sample_path, ProfileData
>>> from heavyedge_features.edge_width import width_type3
>>> with ProfileData(get_sample_path("Prep-Type3.h5")) as data:
...     x = data.x()
...     Y, L, _ = data[0]
>>> sigma = 32.0
>>> b = width_type3(x, Y, L, sigma)
>>> import matplotlib.pyplot as plt
... plt.plot(x, Y, color="gray", alpha=0.5)
... plt.axvline(x[L - 1] - b, color="red", ls="--", label="Edge boundary")
... plt.legend()