def test_iterate_directories():
directory1 = data_path("files/d1")
directory2 = data_path("files/d2")
files = list(iterate_directories((directory1, directory2), "foo"))
assert len(files) == 5
files = list(iterate_directories((directory1, directory2), "foo", "e"))
assert len(files) == 3
files = list(iterate_directories((directory1, directory2), "baz"))
assert len(files) == 0
def test_iterate_files():
directory = data_path("files/d1")
assert len(list(iterate_files(directory, "foo"))) == 3
assert len(list(iterate_files(directory, ".foo"))) == 3
assert len(list(iterate_files(directory, "foo", "e"))) == 2
assert len(list(iterate_files(directory, "foo", "a"))) == 1
assert len(list(iterate_files(directory, "bar"))) == 0
def test_load_dataset():
dataset_path = os.path.abspath(data_path("dataset"))
df = load_dataset(dataset_path, ".jpeg")
assert sorted(df["path"]) == [
os.path.join(dataset_path, f"{p}.jpeg") for p in ("1", "2", "3")
]
df = load_dataset(dataset_path, ".jpeg", "1")
assert sorted(df["path"]) == [os.path.join(dataset_path, "1.jpeg")]
def test_overlay_masks():
masks = [
load_image(data_path(f"segmentation/color_masks/mask-{i}.png"))
for i in range(4)
]
background = np.zeros((100, 100, 3), dtype=np.uint8)
background[:, :] = (50, 50, 50)
check_image_equality(
overlay_masks(background, masks, alpha=1.0),
data_path("segmentation/color_masks/overlay-alpha-1.0.png"),
delta=1,
)
check_image_equality(
overlay_masks(background, masks, alpha=0.5),
data_path("segmentation/color_masks/overlay-alpha-0.5.png"),
delta=1,
)
def test_load_voc_xml():
annotated = voc_to_annotated_image(data_path("example.xml"))
width, height = annotated.width, annotated.height
assert width == 500
assert height == 375
assert annotated.annotations[0].class_name == "dog"
assert annotated.annotations[0].confidence is None
assert annotated.annotations[0].type == AnnotationType.GROUND_TRUTH
assert annotated.annotations[0].bbox.denormalize(
width, height).to_int().as_tuple() == (144, 255, 90, 201)
assert annotated.annotations[1].class_name == "dog"
assert annotated.annotations[1].bbox.denormalize(
width, height).to_int().as_tuple() == (264, 380, 73, 180)
def test_color_bitmap_masks():
mask1 = np.zeros((100, 100), dtype=np.uint8)
mask1[10:30, 20:40] = 255
mask2 = np.zeros((100, 100), dtype=np.uint8)
mask2[35:50, 40:60] = 255
mask3 = np.zeros((100, 100), dtype=np.uint8)
mask3[35:50, 70:90] = 255
mask4 = np.zeros((100, 100), dtype=np.uint8)
mask4[70:95, 10:80] = 255
palette = [(1.0, 0.0, 0.0), (0.0, 1.0, 0.0), (0.0, 0.0, 1.0)]
colored_masks = color_bitmap_masks((mask1, mask2, mask3, mask4), palette)
for (index, mask) in enumerate(colored_masks):
check_image_equality(
mask, data_path(f"segmentation/color_masks/mask-{index}.png"))
def test_load_image_rgb():
img = load_image(data_path("example.jpeg"))
assert img.shape == (375, 500, 3)
assert np.max(img) > 1.0
def test_load_image_normalize():
img = load_image(data_path("example.jpeg"), normalize=True)
assert np.max(img) <= 1.0
def test_load_image_resize():
img = load_image(data_path("example.jpeg"), target_size=(224, 224))
assert img.shape == (224, 224, 3)
def test_load_image_grayscale():
img = load_image(data_path("example.jpeg"), color_mode="grayscale")
assert img.shape == (375, 500)
def test_load_image_bgr():
img2 = load_image(data_path("example.jpeg"), color_mode="bgr")
check_image_equality(cv2.cvtColor(img2, cv2.COLOR_BGR2RGB),
data_path("example.jpeg"))