Skip to content

Commit

Permalink
lint
Browse files Browse the repository at this point in the history
  • Loading branch information
@Atashnezhad
Atashnezhad committed Aug 10, 2023
1 parent be4493a commit 30a5d3c
Show file tree
Hide file tree
Showing 2 changed files with 57 additions and 61 deletions.
4 changes: 3 additions & 1 deletion neural_network_model/transfer_learning.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -608,7 +608,9 @@ def plot_metrics_results(self, **kwargs):
plt.title("History of {}".format(title), fontweight="bold")
plt.legend()
plt.grid()
plt.ylim(0,)
plt.ylim(
0,
)
plt.xlabel("Epoch", fontweight="bold")
plt.ylabel(title, fontweight="bold")
plt.gca().xaxis.set_major_locator(
Expand Down
114 changes: 54 additions & 60 deletions neural_network_model/visual_insights.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -7,10 +7,10 @@
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
from PIL import Image
from skimage import color
from skimage.feature import hessian_matrix, hessian_matrix_eigvals, local_binary_pattern
from skimage.filters import frangi, hessian, meijering, sato, sobel, threshold_multiotsu
from PIL import Image
from sklearn.cluster import KMeans
from tqdm import tqdm

Expand Down Expand Up @@ -429,11 +429,11 @@ def frangi_feature_extraction(
return _frangi_features

def local_binary_pattern(
self,
image,
n_points: int,
radius: int,
method=SUPERVISE_LEARNING_SETTING.FILTERS.LOCAl_BINARY_PATTERN.METHOD
self,
image,
n_points: int,
radius: int,
method=SUPERVISE_LEARNING_SETTING.FILTERS.LOCAl_BINARY_PATTERN.METHOD,
):
"""
Compute the Local Binary Pattern (LBP) representation of an input grayscale image.
Expand Down Expand Up @@ -584,8 +584,8 @@ def lbp_feature_extraction(
return _lbp_features

def multiotsu_threshold(
self,
image,
self,
image,
):
"""
Compute multi-Otsu threshold values for an input image.
Expand Down Expand Up @@ -623,7 +623,8 @@ def multiotsu_threshold_feature_extraction(
plt_show (bool): Whether to display plots (default: False).
plt_log (bool): Whether to use a logarithmic scale for y-axis in histograms (default: False).
figsize (tuple): Size of the displayed figure (default: (10, 10)).
classes (int): Number of classes for multi-Otsu thresholding (default: SUPERVISE_LEARNING_SETTING.FILTERS.MULTIOTSU_THRESHOLD.CLASSES).
classes (int): Number of classes for multi-Otsu thresholding
(default: SUPERVISE_LEARNING_SETTING.FILTERS.MULTIOTSU_THRESHOLD.CLASSES).
Returns:
dict: A dictionary containing histogram features for each channel (R, G, and B).
Expand Down Expand Up @@ -918,7 +919,7 @@ def filter_images(
Returns:
None
"""
"""

cmap = kwargs.get("cmap", "jet")
filter_name = kwargs.get("filter_name", "hessian")
Expand Down Expand Up @@ -1006,9 +1007,9 @@ def filter_images(

if filter_name == "lbp":
for index, row in tqdm(
self.image_df.iterrows(),
total=self.image_df.shape[0],
desc="Filtering images > lbp",
self.image_df.iterrows(),
total=self.image_df.shape[0],
desc="Filtering images > lbp",
):
image_path = row["Filepath"]

Expand All @@ -1019,7 +1020,7 @@ def filter_images(
image_gray,
n_points=SUPERVISE_LEARNING_SETTING.FILTERS.LOCAl_BINARY_PATTERN.NUM_POINTS,
radius=SUPERVISE_LEARNING_SETTING.FILTERS.LOCAl_BINARY_PATTERN.RADIUS,
method=SUPERVISE_LEARNING_SETTING.FILTERS.LOCAl_BINARY_PATTERN.METHOD
method=SUPERVISE_LEARNING_SETTING.FILTERS.LOCAl_BINARY_PATTERN.METHOD,
)

# Get the sub-folder structure from the original image path
Expand Down Expand Up @@ -1047,13 +1048,7 @@ def filter_images(
plt.savefig(filtered_image_path, bbox_inches="tight", pad_inches=0)
plt.close()

def image_segmentation_knn(
self,
image_path,
num_clusters=5,
plt_show=False
):

def image_segmentation_knn(self, image_path, num_clusters=5, plt_show=False):
"""
Apply K-Means clustering for image segmentation and return a colored version of the segmented image.
Expand All @@ -1063,14 +1058,15 @@ def image_segmentation_knn(
Args:
image_path (str): Path to the grayscale image to be segmented.
num_clusters (int, optional): Number of clusters (colors) for image segmentation (default is 5).
plt_show (bool, optional): Whether to show plots of the original grayscale and segmented images (default is False).
plt_show (bool, optional): Whether to show plots of the original grayscale and segmented images
(default is False).
Returns:
np.ndarray: Colored version of the segmented image as a NumPy array.
"""

# Load the grayscale image
gray_image = Image.open(image_path).convert('L')
gray_image = Image.open(image_path).convert("L")

# Convert the grayscale image to a numpy array
gray_array = np.array(gray_image)
Expand All @@ -1097,27 +1093,29 @@ def image_segmentation_knn(
if plt_show:
# Plot the original grayscale image
plt.subplot(1, 2, 1)
plt.imshow(gray_array, cmap='gray')
plt.title('Grayscale Image')
plt.imshow(gray_array, cmap="gray")
plt.title("Grayscale Image")

# Plot the segmented and colored image
plt.subplot(1, 2, 2)
plt.imshow(colored_image, cmap='seismic') # You can choose a colormap you like
plt.title('Segmented and Colored Image')
plt.imshow(
colored_image, cmap="seismic"
) # You can choose a colormap you like
plt.title("Segmented and Colored Image")

plt.tight_layout()
plt.show()

return colored_image

def process_image(
self,
row,
cmap,
dataset_path,
segmentation_dataset_path,
replace_existing,
num_clusters=5
self,
row,
cmap,
dataset_path,
segmentation_dataset_path,
replace_existing,
num_clusters=5,
):
"""
Process and save an image using image segmentation and colormap.
Expand All @@ -1137,12 +1135,15 @@ def process_image(
None
"""


image_path = row["Filepath"]
# Get the sub-folder structure from the original image path
relative_path = Path(image_path).relative_to(dataset_path)
segmented_image_dir = Path(segmentation_dataset_path) / relative_path.parent # Create subdirectory
segmented_image_path = segmented_image_dir / f"{relative_path.stem}_filtered{relative_path.suffix}"
segmented_image_dir = (
Path(segmentation_dataset_path) / relative_path.parent
) # Create subdirectory
segmented_image_path = (
segmented_image_dir / f"{relative_path.stem}_filtered{relative_path.suffix}"
)
# Handle replacing existing images
if not replace_existing:
segmented_image_path = Path(segmentation_dataset_path) / relative_path
Expand All @@ -1159,9 +1160,7 @@ def process_image(
segmented_image_path.parent.mkdir(parents=True, exist_ok=True)

colored_image = self.image_segmentation_knn(
image_path,
num_clusters=num_clusters,
plt_show=False
image_path, num_clusters=num_clusters, plt_show=False
)

# Create a separate figure for each process to avoid conflicts
Expand All @@ -1174,11 +1173,7 @@ def process_image(
plt.savefig(segmented_image_path, bbox_inches="tight", pad_inches=0)
plt.close(fig)

def image_segmentation(
self,
num_clusters=5,
**kwargs
):
def image_segmentation(self, num_clusters=5, **kwargs):
"""
Apply image segmentation to a directory of images using a specified clustering method.
Expand Down Expand Up @@ -1213,16 +1208,19 @@ def image_segmentation(
with multiprocessing.Pool(processes=multiprocessing.cpu_count()) as pool:
results = []
for index, row in tqdm(
self.image_df.iterrows(),
total=self.image_df.shape[0],
desc="Segmentation images > kmean",
self.image_df.iterrows(),
total=self.image_df.shape[0],
desc="Segmentation images > kmean",
):
result = pool.apply_async(
self.process_image,
args=(
row, cmap, dataset_path,
row,
cmap,
dataset_path,
segmentation_dataset_path,
replace_existing, num_clusters
replace_existing,
num_clusters,
),
)
results.append(result)
Expand All @@ -1231,10 +1229,7 @@ def image_segmentation(
for result in results:
result.wait()

def apply_colormap_to_directory(
self,
**kwargs
):
def apply_colormap_to_directory(self, **kwargs):
"""
Apply a specified colormap to a directory of images and save the edited images to a new directory.
Expand All @@ -1247,7 +1242,7 @@ def apply_colormap_to_directory(
Returns:
None
"""
"""

cmap = kwargs.get("cmap", "seismic")
dataset_path = kwargs.get("dataset_path", None)
Expand All @@ -1262,11 +1257,10 @@ def apply_colormap_to_directory(
Path(edited_dataset_path).mkdir(parents=True, exist_ok=True)

for index, row in tqdm(
self.image_df.iterrows(),
total=self.image_df.shape[0],
desc="Editing images > cmap",
self.image_df.iterrows(),
total=self.image_df.shape[0],
desc="Editing images > cmap",
):

image_path = row["Filepath"]
# Get the sub-folder structure from the original image path
relative_path = Path(image_path).relative_to(dataset_path)
Expand Down Expand Up @@ -1375,4 +1369,4 @@ def apply_colormap_to_directory(
dataset_path=dataset_path,
edited_dataset_path=Path(__file__).parent / ".." / "edited_dataset_ad",
replace_existing=False,
)
)

0 comments on commit 30a5d3c

Please sign in to comment.