def show_classification_distributions(labels, data, outliers, data_type, title_flag=True):
"""
Shows distributions of classes after segmentation
Parameters
----------
labels : NumPy Array
matrix of classification per pixel
data : NumPy Array
SPM data supplied by the user
data_type : str
data type corresponding to config.data_info keyword (QNM, AMFM, cAFM)
title_flag : bool
flag for plots to show titles or not
Returns
----------
"""
props = INFO[data_type]["properties"]
if outliers is None:
outliers = 0
unique_labels = slu.get_unique_labels(labels)
num_labels = len(unique_labels)
h, w, c = data.shape
num_plots = 2 * c
num_cols = 2 * NUM_COLS
num_rows = int(np.ceil(num_plots / num_cols))
fig = pyplot.figure(figsize=(20, 15), dpi=80, facecolor="w", edgecolor="k")
cnt = 1
cmap = pyplot.get_cmap("jet", num_labels)
for index_i, i in enumerate(range(c)):
ax_l = pyplot.subplot(num_rows, num_cols, cnt)
cnt += 1
if title_flag:
ax_l.set_title(props[i])
else:
ax_l.set_title(f"PCA component {index_i + 1}")
m = ax_l.imshow(data[:, :, i], aspect="auto")
pyplot.colorbar(m, fraction=0.046, pad=0.04)
ax_r = pyplot.subplot(num_rows, num_cols, cnt)
cnt += 1
ax_r.grid()
ax_r.set_title("Distributions")
for index, j in enumerate(unique_labels):
color_step = num_labels - (index + 1)
labels_wo_outliers = labels * (1 - outliers)
class_data = data[:, :, i][labels_wo_outliers == j]
ax_r.hist(class_data, NUM_BINS, alpha=ALPHA, density=True, color=cmap(color_step))
pyplot.tight_layout()
pyplot.show()
def show_distributions_together(labels, data, data_type, input_cmap):
"""
Shows distributions of classes after segmentation
Parameters
----------
labels : NumPy Array
matrix of classification per pixel
data : NumPy Array
SPM data supplied by the user
data_type : str
data type corresponding to config.data_info keyword (QNM, AMFM, cAFM)
input_cmap : str
specifies which color map to use
Returns
----------
"""
props = INFO[data_type]["properties"]
unique_labels = slu.get_unique_labels(labels)
num_labels = len(unique_labels)
h, w, c = data.shape
num_plots = 2 * c
num_cols = 2 * NUM_COLS
num_rows = int(np.ceil(num_plots / num_cols))
fig = pyplot.figure(figsize=(20, 15), dpi=80, facecolor="w", edgecolor="k")
cnt = 1
cmap = pyplot.get_cmap(input_cmap, num_labels)
for i in range(c):
ax_l = pyplot.subplot(num_rows, num_cols, cnt)
cnt += 1
ax_l.set_title("Segmentation")
ax_l.imshow(labels, alpha=ALPHA, cmap=cmap, aspect="auto")
ax_r = pyplot.subplot(num_rows, num_cols, cnt)
cnt += 1
ax_r.set_title(props[i])
ax_r.grid()
for index, j in enumerate(
unique_labels): # plots mask and distribution per class
color_step = num_labels - (index + 1)
ax_r.hist(data[:, :, i][labels == j],
NUM_BINS,
alpha=ALPHA,
density=True,
color=cmap(color_step))
pyplot.tight_layout()
pyplot.show()
def show_grain_area_distribution(labels, data_type, data_subtype):
"""
Computes a histogram of the number of pixels per label
Parameters
----------
labels : NumPy Array
matrix of classification per pixel
data : NumPy Array
SPM data supplied by the user
data_type : str
data type corresponding to config.data_info keyword (QNM, AMFM, cAFM)
Returns
----------
"""
sample_area = INFO[data_type]["sample_size"][data_subtype] ** 2
unique_labels = slu.get_unique_labels(labels)
grain_areas = sorted([np.sum(labels == l) for l in unique_labels], reverse=True)
percent_grain_areas = grain_areas / np.sum(grain_areas) * 100
physical_grain_areas = percent_grain_areas / 100 * sample_area
pyplot.figure(figsize=(18, 5), dpi=80, facecolor="w", edgecolor="k")
pyplot.subplot(1, 3, 1)
pyplot.plot(np.log10(grain_areas), "ro")
pyplot.plot(np.log10(grain_areas), "b")
pyplot.xlabel("Grain")
pyplot.ylabel("Log number of pixles per grain")
pyplot.title("Log Number of Pixels per Grain")
pyplot.grid()
pyplot.subplot(1, 3, 2)
pyplot.plot(percent_grain_areas, "ro")
pyplot.plot(percent_grain_areas, "b")
pyplot.xlabel("Grain")
pyplot.ylabel("Grain percentage (%)")
pyplot.title("Grain Percentage (%)")
pyplot.grid()
pyplot.subplot(1, 3, 3)
pyplot.plot(physical_grain_areas, "ro")
pyplot.plot(physical_grain_areas, "b")
pyplot.xlabel("Grain")
pyplot.ylabel("Grain area (um2)")
pyplot.title("Grain Area (um2)")
pyplot.grid()
pyplot.show()
def show_classification(labels, data, data_type, input_cmap="jet"):
"""
Shows classification of pixels after segmentation
Parameters
----------
labels : NumPy Array
matrix of classification per pixel
data : NumPy Array
matrix containing user entered SPM data
data_type : str
string designating data type (QNM, AMFM, cAFM)
input_cmap : str
string designating matplotlib colormap to use
Returns
----------
"""
props = INFO[data_type]["properties"]
unique_labels = slu.get_unique_labels(labels)
num_labels = len(unique_labels)
h, w, c = data.shape
num_plots = 2 * c
num_cols = 2 * NUM_COLS
num_rows = int(np.ceil(num_plots / num_cols))
fig = pyplot.figure(figsize=(16, 16), dpi=80, facecolor="w", edgecolor="k")
cnt = 1
cmap = pyplot.get_cmap(input_cmap, num_labels)
for i in range(c):
ax_l = pyplot.subplot(num_rows, num_cols, cnt)
cnt += 1
ax_l.set_title(props[i])
m = ax_l.imshow(data[:, :, i])
pyplot.colorbar(m, fraction=0.046, pad=0.04)
ax_r = pyplot.subplot(num_rows, num_cols, cnt)
cnt += 1
ax_r.set_title("Segmentation")
ax_r.imshow(labels, alpha=ALPHA, cmap=cmap)
colorbar_index(ncolors=num_labels, cmap=cmap)
pyplot.tight_layout()
pyplot.show()
def applies_morphological_cleanup(labels,
bg_label,
morph_radius=5,
morph_min_size=25):
"""
Applies morphological cleanup to labeling assignment per component / grain.
Parameters
---------
labels : NumPy Array
labeling assignment
bg_label : int
GMM background label
morph_radius : int
disk radius for the morphological orperation
morph_min_size : int
remove small objects smaller than this value
Returns
----------
clean_labels : NumPy Array
labeling after morphological cleanup
"""
components = slu.get_unique_labels(labels)
components.remove(bg_label)
# Initialize clean labels as bacdground
clean_labels = (labels == bg_label) * bg_label
for i in components: # non-background gmm components in increasing order
gmm_labels = labels == i
clean_gmm_labels = morphology.binary_opening(
gmm_labels, morphology.disk(morph_radius))
clean_gmm_labels = morphology.remove_small_objects(
clean_gmm_labels, min_size=morph_min_size)
clean_gmm_labels = clean_gmm_labels.astype(np.uint8)
clean_labels[clean_gmm_labels == 1] = i
clean_labels[(clean_gmm_labels == 0) * gmm_labels] = bg_label
return clean_labels
def show_classification_correlation(labels, data, data_type, title_flag=True, sample_flag=True):
"""
Plots the correlation of data properties after classification
Parameters
----------
data : NumPy Array
SPM data supplied by the user
data_type : str
data type corresponding to config.data_info keyword (QNM, AMFM, cAFM)
title_flag : bool
flag for plots to show titles or not
sample_flag : bool
flag to sample points for visualization
data : NumPy Array
SPM data supplied by the user
data_type : str
data type corresponding to config.data_info keyword (QNM, AMFM, cAFM)
outliers : NumPy Array
boolean, 2D array of outlier flags (1's) for functions to pass over
Returns
----------
"""
props = INFO[data_type]["properties"]
unique_labels = slu.get_unique_labels(labels)
num_labels = len(unique_labels)
h, w, c = data.shape
num_plots = np.sum(np.arange(c))
num_cols = 2 * NUM_COLS
num_rows = int(np.ceil(num_plots / num_cols))
fig = pyplot.figure(figsize=(20, 18), dpi=80, facecolor="w", edgecolor="k")
cmap = pyplot.get_cmap("jet", num_labels)
cnt = 1
for index_i, i in enumerate(range(c)):
for index_j, j in enumerate(range(c)):
if index_j <= index_i:
continue
ax = pyplot.subplot(num_rows, num_cols, cnt)
ax.grid()
if title_flag:
ax.set_xlabel(props[index_i])
ax.set_ylabel(props[index_j])
else:
ax.set_xlabel(f"PCA component {index_i + 1}")
ax.set_ylabel(f"PCA component {index_j + 1}")
cnt += 1
title = f"Correlation: "
for index, l in enumerate(unique_labels):
data_i = data[:, :, index_i][labels == l]
data_i /= np.max(np.abs(data_i))
data_j = data[:, :, index_j][labels == l]
data_j /= np.max(np.abs(data_j))
corr = np.corrcoef(data_i, data_j)[0, 1]
if sample_flag:
indices = np.arange(data_i.shape[0])
sample_size = min(1000, len(indices))
sampling = np.random.choice(indices, size=sample_size, replace=False)
data_i = np.take(data_i, sampling)
data_j = np.take(data_j, sampling)
color_step = num_labels - (index + 1)
ax.scatter(data_i, data_j, color=cmap(color_step), alpha=0.2)
title += f"{index}: {np.round(corr, 3)} "
ax.set_title(title)
pyplot.tight_layout()
pyplot.show()