def test_create_charset_dir_correct_path_nosplit(self):
"""Check if the path to the new directory has been returned."""
prefix = tempfile.mkdtemp()
gen = CharImageGenerator(out_dir=prefix)
charset_dir, = gen.create_charset_dir(self.TEST_CHARSET,
test_train_split=False,
create_parent_dir=True)
self.assertTrue(charset_dir.startswith(prefix))
def test_create_charset_dir_nosplit(self):
"""Check if the new directory contains subdirectories matching the test_charset."""
prefix = tempfile.mkdtemp()
gen = CharImageGenerator(out_dir=prefix)
charset_dir, = gen.create_charset_dir(self.TEST_CHARSET,
test_train_split=False,
create_parent_dir=True)
charset_ascii = set(ord(c) for c in self.TEST_CHARSET)
subdirs = set(int(dir_name) for dir_name in os.listdir(charset_dir))
self.assertEqual(subdirs, charset_ascii)
def test_generate_char_images(self):
import types
prefix = tempfile.mkdtemp()
n_samples = 3
gen = CharImageGenerator(out_dir=prefix, charset=self.TEST_CHARSET)
image_generator = gen.generate_char_images(n_samples=n_samples,
augment=True)
self.assertIsInstance(image_generator, types.GeneratorType)
expected_img_count = len(self.TEST_CHARSET) * n_samples
img_count = sum(1 for _ in image_generator)
self.assertEqual(img_count,
expected_img_count,
msg="Number of generated images"
" does not match the expected value.")
def test_load_charset(self):
fd, f_path = tempfile.mkstemp()
for char in self.TEST_CHARSET:
os.write(fd, "{}\n".format(char).encode())
os.close(fd)
loaded_charset = CharImageGenerator.load_char_set(f_path)
self.assertSequenceEqual(self.TEST_CHARSET, loaded_charset)
def test_create_and_save_charsets_default(self):
prefix = tempfile.mkdtemp()
n_samples = 3
gen = CharImageGenerator(out_dir=prefix, charset=self.TEST_CHARSET)
gen.create_and_save_charsets(test_train_split=True,
n_samples=n_samples,
augment=True)
# check that the dataset has been split into the test and train directory
self.assertEqual(set(os.listdir(prefix)), {'test_data', 'train_data'})
# and that they are not empty
expected_img_count = len(self.TEST_CHARSET) * n_samples
img_count = 0
for root, _, walkfiles in os.walk(prefix):
img_count += len(walkfiles)
self.assertEqual(img_count,
expected_img_count,
msg="Number of created images"
" does not match the expected value.")
def main():
# Initialize colored output
colorama.init()
parser = argparse.ArgumentParser()
parser.add_argument(
'--charset',
help=
"Path to a file containing characters to be turned into images, separated by newlines."
)
parser.add_argument(
'--font-dir',
help=
"Path to a directory containing font files (obtained by scraper), separated by newlines."
)
parser.add_argument(
'-p',
'--prefix',
help=
"If provided, generated directory will be prefixed with the value of `prefix`."
)
args = parser.parse_args()
gen = CharImageGenerator.load(charset_path=args.charset,
fonts_path=args.font_dir,
out_dir=args.prefix)
print(f"{colorama.Fore.YELLOW}Creating sprite sheets ...")
gen.create_sprites(
) # Generates sprite sheets as a preview of fonts - no augmentation performed
print(
f"{colorama.Fore.GREEN}Sprite sheets have been created successfully.")
print(f"{colorama.Fore.YELLOW}Generating character images ...")
gen.create_and_save_charsets(
test_train_split=True
) # Also creates default charset dir if not existent
print(f"{colorama.Fore.GREEN}Image generation completed successfully.")
def test_char_img_gen_init(self):
char_gen = CharImageGenerator()
self.assertIsNotNone(char_gen)