Added new test, cleaned up code

tests/helpers/test_rotation_scan.py

@@ -4,40 +4,88 @@ import PIL
 import os
 import math
 import numpy as np
-from zesje.helpers import image_helper,pdf_helper
+from zesje.helpers import image_helper
+
+
+# Helper functions
+# Given 2 keypoints, calculates the distance between them
+def distance(keyp1, keyp2):
+    return math.hypot(keyp1.pt[0] - keyp2.pt[0], keyp1.pt[1] - keyp2.pt[1])
+
+
+# Given a name of a exam image and the location it is stored, retrieves the
+# image and converts it to binary image
+def generate_binary_image(name, datadir):
+    pdf_path = os.path.join(datadir, 'scanned_pdfs', f'{name}')
+    pil_im = PIL.Image.open(pdf_path)
+    opencv_im = cv2.cvtColor(np.array(pil_im), cv2.COLOR_RGB2BGR)
+    _, bin_im = cv2.threshold(opencv_im, 150, 255, cv2.THRESH_BINARY)
+    return bin_im
 
 # Tests
 
-@pytest.mark.parametrize('test_input1, test_input2, expected', [
-    (cv2.KeyPoint(1337,69,0), cv2.KeyPoint(9001,69,0), 0),
-    (cv2.KeyPoint(0,100,0), cv2.KeyPoint(0,1000,0), 90),
-    (cv2.KeyPoint(25,25,0), cv2.KeyPoint(50,50,0), -45),
-    (cv2.KeyPoint(25,25,0), cv2.KeyPoint(50,0,0), 45)],
-    ids = ['Same horizontal line', 'Same vertical line', 'Negative angle', 'Positive angle']
-)
 
+# Tests whether the output of calc angle is correct
+@pytest.mark.parametrize('test_input1, test_input2, expected', [
+    (cv2.KeyPoint(1337, 69, 0), cv2.KeyPoint(9001, 69, 0), 0),
+    (cv2.KeyPoint(0, 100, 0), cv2.KeyPoint(0, 1000, 0), 90),
+    (cv2.KeyPoint(25, 25, 0), cv2.KeyPoint(50, 50, 0), -45),
+    (cv2.KeyPoint(25, 25, 0), cv2.KeyPoint(50, 0, 0), 45)],
+    ids=['Same horizontal line', 'Same vertical line', 'Negative angle',
+         'Positive angle'])
 def test_calc_angle(test_input1, test_input2, expected):
-    assert math.isclose(image_helper.calc_angle(test_input1, test_input2), expected, abs_tol=0.1)
+    assert math.isclose(image_helper.calc_angle(test_input1, test_input2),
+                        expected, abs_tol=0.1)
 
 
-@pytest.mark.parametrize('name',
-    os.listdir(os.path.join('tests','data','scanned_pdfs')),
-    ids = os.listdir(os.path.join('tests','data','scanned_pdfs')))
+# Tests whether the amount of cornermakers is enough to calculate the angle and
+# whether it is lower than 5 as we only add 4 corner markers per page.
+@pytest.mark.parametrize('name', os.listdir(
+                                 os.path.join('tests',
+                                              'data', 'scanned_pdfs')),
+                         ids=os.listdir(
+                            os.path.join('tests', 'data', 'scanned_pdfs')))
+def test_detect_enough_cornermarkers(name, datadir):
+    bin_im = generate_binary_image(name, datadir)
+    keypoints = image_helper.find_corner_marker_keypoints(bin_im)
+    assert(len(keypoints) >= 2 & len(keypoints) <= 4)
 
-def test_detect_enough_cornermarkers(name,datadir):
-    pdf_path = os.path.join(datadir,'scanned_pdfs', f'{name}')
 
-    pil_im = PIL.Image.open(pdf_path)
+# Tests whether the detected keypoints are actually corner markers.
+# This is done by checking whether they are close enough to the corner
+# of the image. Only A4 is considered as there is no test data yet for
+# US letter size.
+@pytest.mark.parametrize('name', os.listdir(
+                                 os.path.join('tests',
+                                              'data', 'scanned_pdfs')),
+                         ids=os.listdir(
+                            os.path.join('tests', 'data', 'scanned_pdfs')))
+def test_detect_valid_cornermarkers(name, datadir):
+    bin_im = generate_binary_image(name, datadir)
+    keypoints = image_helper.find_corner_marker_keypoints(bin_im)
 
-    opencv_im = cv2.cvtColor(np.array(pil_im), cv2.COLOR_RGB2BGR)
+    h, w, *_ = bin_im.shape
+    (xmm, ymm) = (210, 297)
+    (xcorner, ycorner) = (round(30 * w / xmm), round(30 * h / ymm))
+    maxdist = math.hypot(xcorner, ycorner)
 
-    _, bin_im = cv2.threshold(opencv_im, 150, 255, cv2.THRESH_BINARY)
+    cornerlist = [cv2.KeyPoint(0, 0, 0), cv2.KeyPoint(w, 0, 0),
+                  cv2.KeyPoint(0, h, 0), cv2.KeyPoint(w, h, 0)]
 
-    keypoints = image_helper.find_corner_marker_keypoints(bin_im)
+    # Checks whether there aren't multiple keypoints in the same corner.
+    # If there is, one of those probably isn't a corner marker.
+    result = np.array([0, 0, 0, 0])
+    for detected_keypoint in keypoints:
+        distlist = np.array([distance(detected_keypoint, corner_keypoint)
+                            for corner_keypoint in cornerlist])
+        binlist = distlist < maxdist
+        result = binlist + result
+
+    assert(sum(result > 1) == 0)
 
-    assert(len(keypoints) >= 2 & len(keypoints) <= 4)
 
 # Untested:
 #
 # Other parts of the rotation function are not tested due to it being opencv
 # functions.
+#