Compatibility to Python up to 3.14, two bugfixes, hardening

kalmus/command_line_generator.py

@@ -71,3 +71,7 @@ def main(args=sys.argv[1:]):
                                        rescale_frames_factor=args["rescale_frame_factor"])
     barcode = barcode_generator.get_barcode()
     barcode.save_as_json(filename=args["output_path"])
+
+
+if __name__ == "__main__":
+    main()

kalmus/command_line_gui.py

@@ -33,3 +33,7 @@ def main(args=sys.argv[1:]):
 
     # Use the default barcode and the barcode generator to Instantiate the Main window of the kalmus software (GUI)
     MainWindow(barcode_tmp, barcode_gn, dpi=args["dpi"])
+
+
+if __name__ == "__main__":
+    main()

kalmus/tkinter_windows/SaveImageWindow.py

@@ -125,6 +125,13 @@ def save_image(self):
             showerror("File Name is Not Given", "Please specify the path to the saved image.")
             return
 
+        # On Linux/X11 the file dialog does not auto-append the filtered extension, and a
+        # user typing the path by hand may also omit it. Default to .jpg so the call into
+        # matplotlib/PIL has something to dispatch on.
+        _, ext = os.path.splitext(filename)
+        if ext == "":
+            filename += ".jpg"
+
         # Get which barcode to save
         if self.barcode_option.get() == "Barcode 1":
             barcode = self.barcode_1.get_barcode().astype("uint8")
@@ -137,11 +144,17 @@ def save_image(self):
         barcode = cv2.resize(barcode, dsize=(int(self.resize_x_entry.get()), int(self.resize_y_entry.get())),
                              interpolation=cv2.INTER_NEAREST)
 
-        # Save the barcode with desirable color map based on its barcode type
-        if barcode_type == "Color":
-            plt.imsave(filename, barcode)
-        else:
-            plt.imsave(filename, barcode, cmap="gray")
+        # Save the barcode with desirable color map based on its barcode type. Surface any
+        # save error (unsupported extension, permission denied, ...) as a dialog instead of
+        # letting it tear up the Tk callback.
+        try:
+            if barcode_type == "Color":
+                plt.imsave(filename, barcode)
+            else:
+                plt.imsave(filename, barcode, cmap="gray")
+        except (ValueError, OSError) as exc:
+            showerror("Could Not Save Image", "Failed to save image to {:s}\n\n{:s}".format(filename, str(exc)))
+            return
 
         # Quit the window
         self.window.destroy()

kalmus/tkinter_windows/gui_utils.py

@@ -285,8 +285,9 @@ def update_hist(barcode, ax, bin_step=5):
         # Paint each bin with its corresponding color in hue
         paint_hue_hist(bin_step, patches)
     else:
-        # If the barcode type is brightness
-        N, bins, patches = ax.hist(barcode.brightness[:, 0], bins=(np.arange(0, 256, bin_step)))
+        # If the barcode type is brightness. `brightness` is stored 1-D (one scalar per
+        # sampled frame), so flatten defensively rather than indexing a phantom 2nd axis.
+        N, bins, patches = ax.hist(barcode.brightness.ravel(), bins=(np.arange(0, 256, bin_step)))
 
         # Then paint each bin with its brightness intensity
         paint_gray_hist(bin_step, patches, opacity=0.9)

kalmus/utils/artist.py

@@ -1,5 +1,6 @@
 """ Utility artist """
 
+import inspect
 import numpy as np
 import cv2
 import csv
@@ -11,6 +12,17 @@
 import scipy.stats as stats
 
 
+# skimage>=0.26 deprecates `min_size` (removes objects with size < N) in favour of
+# `max_size` (removes objects with size <= N). The equivalence is `max_size = N - 1`.
+# This shim picks the right kwarg at import time so we work across both APIs.
+if "max_size" in inspect.signature(remove_small_objects).parameters:
+    def _remove_objects_smaller_than(arr, threshold):
+        return remove_small_objects(arr, max_size=threshold - 1)
+else:
+    def _remove_objects_smaller_than(arr, threshold):
+        return remove_small_objects(arr, min_size=threshold)
+
+
 def compute_dominant_color(image, n_clusters=3, max_iter=10, threshold_error=1.0, attempts=10):
     """
     Compute the dominant color of an input image using the kmeans clustering. The centers of the
@@ -109,8 +121,11 @@ def compute_mode_color(image, bin_size=10):
     """
     image = flatten_image(image)
     image = image // bin_size
-    mode_color, counts = stats.mode(image, axis=0)
-    return (mode_color[0] * bin_size), counts
+    # scipy>=1.11 changed stats.mode default to keepdims=False; pin the shape
+    # explicitly so callers that expect (channels,) keep working across versions.
+    mode_result = stats.mode(image, axis=0, keepdims=False)
+    mode_color, counts = mode_result.mode, mode_result.count
+    return mode_color * bin_size, counts
 
 
 def compute_mean_color(image):
@@ -362,7 +377,7 @@ def watershed_segmentation(image, minimum_segment_size=0.0004, base_ratio=0.5, d
     segment_min_size = int(num_of_pixels * minimum_segment_size)
     # find continuous region which marked by gradient lower than critical gradient
     markers = rank.gradient(denoised, disk(marker_disk_size)) < critical
-    markers = ndimage.label(remove_small_objects(markers, segment_min_size))[0]
+    markers = ndimage.label(_remove_objects_smaller_than(markers, segment_min_size))[0]
 
     # process the watershed transformation
     regions = watershed(gradient, markers)

kalmus/utils/measure_utils.py

@@ -1,11 +1,26 @@
 """ Image Comparison Utility """
 
-import Bio.pairwise2 as sequence_align
 import numpy as np
+from Bio.Align import PairwiseAligner
 from skimage.color import rgb2hsv
 from skimage.metrics import mean_squared_error, structural_similarity
 
 
+def _build_aligner(mode, match_score, mismatch_penal, gap_penal, extending_gap_penal):
+    """Construct a PairwiseAligner mirroring the pairwise2.align.{globalms,localms} scoring model.
+
+    pairwise2 applied the same gap penalty to both terminal and internal gaps; assigning
+    ``open_gap_score`` / ``extend_gap_score`` on PairwiseAligner does the same.
+    """
+    aligner = PairwiseAligner()
+    aligner.mode = mode
+    aligner.match_score = match_score
+    aligner.mismatch_score = mismatch_penal
+    aligner.open_gap_score = gap_penal
+    aligner.extend_gap_score = extending_gap_penal
+    return aligner
+
+
 def nrmse_similarity(image_1, image_2, norm_mode="Min max"):
     """
     Normalized root mean squared error (NRMSE).
@@ -61,10 +76,19 @@ def ssim_similarity(image_1, image_2, window_size=None):
     image_1 = image_1.astype("float64")
     image_2 = image_2.astype("float64")
 
+    # skimage requires explicit `data_range` for floating-point inputs since 0.21.
+    # Derive it from the joint value range so behaviour matches the prior int default.
+    data_range = max(image_1.max(), image_2.max()) - min(image_1.min(), image_2.min())
+    if data_range == 0:
+        data_range = 1.0
+
+    # scikit-image>=0.21 dropped the `multichannel` kwarg in favour of `channel_axis`.
     if len(image_1.shape) == 2:
-        score = structural_similarity(image_1, image_2, win_size=window_size, multichannel=False)
+        score = structural_similarity(image_1, image_2, win_size=window_size,
+                                      channel_axis=None, data_range=data_range)
     elif len(image_1.shape) > 2:
-        score = structural_similarity(image_1, image_2, win_size=window_size, multichannel=True)
+        score = structural_similarity(image_1, image_2, win_size=window_size,
+                                      channel_axis=-1, data_range=data_range)
 
     # Renormalize [-1, 1] score to [0, 1] range
     score += 1
@@ -264,12 +288,11 @@ def compare_needleman_wunsch(str_barcode_1, str_barcode_2, local_sequence_size=2
     """
     assert len(str_barcode_1) == len(str_barcode_2), "The lengths of two barcodes have to be identical"
 
+    aligner = _build_aligner("global", match_score, mismatch_penal, gap_penal, extending_gap_penal)
     scores = 0
     for start_point in range(0, len(str_barcode_1), local_sequence_size):
-        scores += sequence_align.align.globalms(str_barcode_1[start_point:start_point + local_sequence_size],
-                                                str_barcode_2[start_point:start_point + local_sequence_size],
-                                                match_score, mismatch_penal, gap_penal, extending_gap_penal,
-                                                score_only=True)
+        scores += aligner.score(str_barcode_1[start_point:start_point + local_sequence_size],
+                                str_barcode_2[start_point:start_point + local_sequence_size])
 
     if normalized:
         denom = len(str_barcode_1) * match_score
@@ -307,12 +330,11 @@ def compare_smith_waterman(str_barcode_1, str_barcode_2, local_sequence_size=200
     """
     assert len(str_barcode_1) == len(str_barcode_2), "The lengths of two barcodes have to be identical"
 
+    aligner = _build_aligner("local", match_score, mismatch_penal, gap_penal, extending_gap_penal)
     scores = 0
     for start_point in range(0, len(str_barcode_1), local_sequence_size):
-        scores += sequence_align.align.localms(str_barcode_1[start_point:start_point + local_sequence_size],
-                                               str_barcode_2[start_point:start_point + local_sequence_size],
-                                               match_score, mismatch_penal, gap_penal, extending_gap_penal,
-                                               score_only=True)
+        scores += aligner.score(str_barcode_1[start_point:start_point + local_sequence_size],
+                                str_barcode_2[start_point:start_point + local_sequence_size])
 
     if normalized:
         denom = len(str_barcode_1) * match_score

requirements.txt

@@ -1,61 +1,11 @@
-attrs==19.1.0
-backcall==0.1.0
-biopython==1.78
-bleach>=3.3.0
-certifi==2020.12.5
-cycler==0.10.0
-decorator==4.4.0
-defusedxml==0.6.0
-entrypoints==0.3
-imageio==2.9.0
-ipykernel==5.1.1
-ipython>=7.16.3
-ipython-genutils==0.2.0
-ipywidgets==7.5.0
-jedi==0.14.1
-Jinja2>=2.11.3
-joblib==1.0.0
-jsonschema==3.0.1
-jupyter==1.0.0
-jupyter-console==6.0.0
-kiwisolver==1.3.1
-MarkupSafe==1.1.1
-matplotlib==3.2.2
-mistune==0.8.4
-mpmath==1.1.0
-nbconvert==5.5.0
-nbformat==4.4.0
-networkx==2.5
-nose==1.3.7
-notebook>=6.1.5
-numpy>=1.21
-opencv-python==4.4.0.46
-pandocfilters==1.4.2
-parso==0.5.1
-pickleshare==0.7.5
-Pillow>=8.1.1
-prometheus-client==0.7.1
-prompt-toolkit==2.0.9
-Pygments>=2.7.4
-pyparsing==2.4.1.1
-pyrsistent==0.15.4
-python-dateutil==2.8.0
-pytz==2019.1
-PyWavelets==1.1.1
-pywinpty==0.5.5
-qtconsole==4.5.2
-scikit-image>=0.16.2
-scikit-learn==0.24.0
-scipy==1.5.4
-Send2Trash==1.5.0
-six==1.12.0
-sympy==1.4
-terminado==0.8.2
-testpath==0.4.2
-threadpoolctl==2.1.0
-tifffile==2020.12.8
-traitlets==4.3.2
-wcwidth==0.1.7
-webencodings==0.5.1
-widgetsnbextension==3.5.0
-wincertstore==0.2
+# Runtime dependencies for kalmus. Mirrors setup.py install_requires.
+# For the optional notebook/dev environment, install jupyter/ipykernel/nbconvert separately.
+biopython>=1.80
+kiwisolver>=1.4
+matplotlib>=3.5
+numpy>=1.23
+opencv-python-headless>=4.6
+pandas>=1.5
+scikit-image>=0.21
+scikit-learn>=1.1
+scipy>=1.11

setup.py

@@ -21,8 +21,11 @@
       classifiers=[
           "License :: OSI Approved :: MIT License",
           "Programming Language :: Python :: 3",
-          "Programming Language :: Python :: 3.7",
-          "Programming Language :: Python :: 3.8",
+          "Programming Language :: Python :: 3.10",
+          "Programming Language :: Python :: 3.11",
+          "Programming Language :: Python :: 3.12",
+          "Programming Language :: Python :: 3.13",
+          "Programming Language :: Python :: 3.14",
           "Development Status :: 5 - Production/Stable",
           "Intended Audience :: Education",
           "Intended Audience :: Science/Research",
@@ -32,9 +35,21 @@
           "Topic :: Software Development :: User Interfaces"
       ],
       packages=find_packages(),
-      python_requires='>=3.7',
-      install_requires=['numpy', 'opencv-python', 'scikit-image>=0.16.2', 'matplotlib>=3.2.2',
-                        'scikit-learn', 'biopython', 'scipy', 'kiwisolver>=1.3.1', 'pandas'],
+      python_requires='>=3.10',
+      install_requires=[
+          'numpy>=1.23',
+          # headless variant: same cv2 API, no X/GL system libs required.
+          # kalmus uses cv2 only for image-processing (VideoCapture, cvtColor,
+          # kmeans, grabCut, ...); the GUI is tkinter, never cv2.imshow.
+          'opencv-python-headless>=4.6',
+          'scikit-image>=0.21',  # channel_axis kwarg replaces multichannel
+          'matplotlib>=3.5',
+          'scikit-learn>=1.1',
+          'biopython>=1.80',     # PairwiseAligner is the supported aligner
+          'scipy>=1.11',         # stats.mode keepdims default change handled
+          'kiwisolver>=1.4',
+          'pandas>=1.5',
+      ],
       entry_points={
           'console_scripts': ['kalmus-gui=kalmus.command_line_gui:main',
                               'kalmus-generator=kalmus.command_line_generator:main'],

tests/test_artist.py

@@ -40,7 +40,7 @@ def test_compute_mode_color(get_test_color_image):
     image = get_test_color_image
     color, count = artist.compute_mode_color(get_test_color_image)
     assert color.shape == (3,)
-    assert count.shape == (1, 3)
+    assert count.shape == (3,)
     assert np.all(count < image.size)