def test_shape_outline(self):
"""Test the shape:outline feature."""
path = IMAGES_RECTANGLE[2]
k = 20
# Object width and height.
w, h = (300, 100)
im_path = os.path.join(self.base_dir, path)
img = cv2.imread(im_path)
if img == None or img.size == 0:
raise SystemError("Failed to read %s" % im_path)
# Perform segmentation.
mask = grabcut(img, 1, None, 1)
bin_mask = np.where((mask == cv2.GC_FGD) + (mask == cv2.GC_PR_FGD),
255, 0).astype('uint8')
# Obtain contours (all points) from the mask.
contour = ft.get_largest_contour(bin_mask.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_NONE)
# Get the outline.
outline = ft.shape_outline(contour, k)
# Assert the length of the return value.
self.assertEqual(len(outline), k)
# Assert the shape values.
for x, y in outline:
delta_x = x[0] - x[1]
delta_y = y[0] - y[1]
self.assertLess(error(delta_y, h, mape), 0.0001)
self.assertLess(error(delta_x, w, mape), 0.0001)
def test_shape_outline(self):
"""Test the shape:outline feature."""
path = IMAGES_RECTANGLE[2]
k = 20
# Object width and height.
w, h = (300, 100)
im_path = os.path.join(self.base_dir, path)
img = cv2.imread(im_path)
if img == None or img.size == 0:
raise SystemError("Failed to read %s" % im_path)
# Perform segmentation.
mask = grabcut(img, 1, None, 1)
bin_mask = np.where((mask==cv2.GC_FGD) + (mask==cv2.GC_PR_FGD), 255, 0).astype('uint8')
# Obtain contours (all points) from the mask.
contour = ft.get_largest_contour(bin_mask.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
# Get the outline.
outline = ft.shape_outline(contour, k)
# Assert the length of the return value.
self.assertEqual( len(outline), k )
# Assert the shape values.
for x, y in outline:
delta_x = x[0] - x[1]
delta_y = y[0] - y[1]
self.assertLess( error(delta_y, h, mape), 0.0001 )
self.assertLess( error(delta_x, w, mape), 0.0001 )
def test_shape_360(self):
"""Test the shape:360 feature.
Extracts the shape from two slightly rotated versions of the same
image. Then the medium square error between the two extracted shapes is
calculated and checked.
"""
max_error = 0.05
shape = []
for i, path in enumerate(IMAGES_ERYCINA):
im_path = os.path.join(self.base_dir, path)
img = cv2.imread(im_path)
if img == None or img.size == 0:
raise SystemError("Failed to read %s" % im_path)
# Resize the image if it is larger then the threshold.
img = scale_max_perimeter(img, MAX_SIZE)
# Perform segmentation.
mask = grabcut(img, 5, None, 1)
# Create a binary mask. Foreground is made white, background black.
bin_mask = np.where((mask == cv2.GC_FGD) + (mask == cv2.GC_PR_FGD),
255, 0).astype('uint8')
# Obtain contours (all points) from the mask.
contour = ft.get_largest_contour(bin_mask, cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_NONE)
# Fit an ellipse on the contour to get the rotation angle.
box = cv2.fitEllipse(contour)
rotation = int(box[2])
# Extract the shape.
intersects, center = ft.shape_360(contour, rotation)
# For each angle save the mean distance from center to contour and
# the standard deviation for the distances.
means = []
sds = []
for angle in range(0, 360):
distances = []
for p in intersects[angle]:
d = ft.point_dist(center, p)
distances.append(d)
if len(distances) == 0:
mean = 0
sd = 0
else:
mean = np.mean(distances, dtype=np.float32)
if len(distances) > 1:
sd = np.std(distances, ddof=1, dtype=np.float32)
else:
sd = 0
means.append(mean)
sds.append(sd)
# Normalize and save result.
means = cv2.normalize(np.array(means), None, -1, 1,
cv2.NORM_MINMAX)
sds = cv2.normalize(np.array(sds), None, -1, 1, cv2.NORM_MINMAX)
shape.append([means, sds])
# Check the medium square error for the means.
self.assertLess(error(shape[0][0].ravel(), shape[1][0].ravel(), mse),
max_error)
# Check the medium square error for the standard deviations.
self.assertLess(error(shape[0][1].ravel(), shape[1][1].ravel(), mse),
max_error)
def test_shape_360(self):
"""Test the shape:360 feature.
Extracts the shape from two slightly rotated versions of the same
image. Then the medium square error between the two extracted shapes is
calculated and checked.
"""
max_error = 0.05
shape = []
for i, path in enumerate(IMAGES_ERYCINA):
im_path = os.path.join(self.base_dir, path)
img = cv2.imread(im_path)
if img == None or img.size == 0:
raise SystemError("Failed to read %s" % im_path)
# Resize the image if it is larger then the threshold.
img = scale_max_perimeter(img, MAX_SIZE)
# Perform segmentation.
mask = grabcut(img, 5, None, 1)
# Create a binary mask. Foreground is made white, background black.
bin_mask = np.where((mask==cv2.GC_FGD) + (mask==cv2.GC_PR_FGD), 255, 0).astype('uint8')
# Obtain contours (all points) from the mask.
contour = ft.get_largest_contour(bin_mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
# Fit an ellipse on the contour to get the rotation angle.
box = cv2.fitEllipse(contour)
rotation = int(box[2])
# Extract the shape.
intersects, center = ft.shape_360(contour, rotation)
# For each angle save the mean distance from center to contour and
# the standard deviation for the distances.
means = []
sds = []
for angle in range(0, 360):
distances = []
for p in intersects[angle]:
d = ft.point_dist(center, p)
distances.append(d)
if len(distances) == 0:
mean = 0
sd = 0
else:
mean = np.mean(distances, dtype=np.float32)
if len(distances) > 1:
sd = np.std(distances, ddof=1, dtype=np.float32)
else:
sd = 0
means.append(mean)
sds.append(sd)
# Normalize and save result.
means = cv2.normalize(np.array(means), None, -1, 1, cv2.NORM_MINMAX)
sds = cv2.normalize(np.array(sds), None, -1, 1, cv2.NORM_MINMAX)
shape.append([means, sds])
# Check the medium square error for the means.
self.assertLess(error(shape[0][0].ravel(), shape[1][0].ravel(), mse), max_error)
# Check the medium square error for the standard deviations.
self.assertLess(error(shape[0][1].ravel(), shape[1][1].ravel(), mse), max_error)
