Refactoring

cv_analysis/figure_detection/figures.py

@@ -1,7 +1,7 @@
 import cv2
 import numpy as np
 
-from cv_analysis.utils.common import find_contours
+from cv_analysis.utils.common import find_contours_and_hierarchies
 
 
 def detect_large_coherent_structures(image: np.array):
@@ -28,6 +28,6 @@ def detect_large_coherent_structures(image: np.array):
     # FIXME: Parameterize via factory
     close = cv2.morphologyEx(dilate, cv2.MORPH_CLOSE, close_kernel, iterations=1)  # TODO: Tweak iterations
 
-    contours, _ = find_contours(close)
+    contours, _ = find_contours_and_hierarchies(close)
 
     return contours

cv_analysis/layout_parsing.py

@@ -5,8 +5,9 @@ import cv2
 import numpy as np
 from funcy import compose, rcompose, lkeep
 
+from cv_analysis.utils import lstarkeep
 from cv_analysis.utils.common import (
-    find_contours,
+    find_contours_and_hierarchies,
     dilate_page_components,
     normalize_to_gray_scale,
     threshold_image,
@@ -20,6 +21,14 @@ from cv_analysis.utils.rectangle import Rectangle
 
 
 def parse_layout(image: np.array) -> List[Rectangle]:
+    """Parse the layout of a page.
+
+    Args:
+        image: Image of the page.
+
+    Returns:
+        List of rectangles representing the layout of the page as identified page elements.
+    """
     rectangles = rcompose(
         find_segments,
         remove_included,
@@ -31,6 +40,7 @@ def parse_layout(image: np.array) -> List[Rectangle]:
 
 
 def find_segments(image: np.ndarray) -> List[Rectangle]:
+    """Find segments in a page. Segments are structural elements of a page, such as text blocks, tables, etc."""
     rectangles = rcompose(
         prepare_for_initial_detection,
         __find_segments,
@@ -49,17 +59,18 @@ def __find_segments(image: np.ndarray) -> List[Rectangle]:
     def to_rectangle_if_valid(contour, hierarchy):
         return contour_to_rectangle(contour) if is_likely_segment(contour) and has_no_parent(hierarchy) else None
 
-    rectangles = lkeep(map(to_rectangle_if_valid, *find_contours(image)))
+    rectangles = lstarkeep(to_rectangle_if_valid, zip(*find_contours_and_hierarchies(image)))
 
     return rectangles
 
 
 def prepare_for_meta_detection(image: np.ndarray, rectangles: Iterable[Rectangle]) -> np.ndarray:
-
-    image = fill_rectangles(image, rectangles)
-    image = threshold_image(image)
-    image = invert_image(image)
-    image = normalize_to_gray_scale(image)
+    image = rcompose(
+        fill_rectangles,
+        threshold_image,
+        invert_image,
+        normalize_to_gray_scale,
+    )(image, rectangles)
 
     return image
 

cv_analysis/utils/common.py

@@ -8,7 +8,7 @@ from funcy import first
 from cv_analysis.utils.rectangle import Rectangle
 
 
-def find_contours(image):
+def find_contours_and_hierarchies(image):
     contours, hierarchies = cv2.findContours(image, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
     return contours, first(hierarchies) if hierarchies is not None else None
 

cv_analysis/utils/utils.py

@@ -1,7 +1,7 @@
 from __future__ import annotations
 
 import cv2
-from funcy import first, iterate
+from funcy import first, iterate, keep
 from numpy import generic
 
 
@@ -30,12 +30,20 @@ def lift(fn):
 
 
 def star(fn):
-    def starred(*args):
+    def starred(args):
         return fn(*args)
 
     return starred
 
 
+def lstarkeep(fn, coll):
+    return list(starkeep(fn, coll))
+
+
+def starkeep(fn, coll):
+    yield from keep(star(fn), coll)
+
+
 def until(cond, func, *args, **kwargs):
     return first(filter(cond, iterate(func, *args, **kwargs)))