def create_candidate_data(candidate: str) -> dict:
data = { }
for fontfamily in fontfamilies:
for fontsize in fontsizes:
path = os.path.join(fontfamily, str(fontsize))
img = image.create_image(candidate, fontsize, fontfamily)
blurred_img = image.blur(img, kernel[fontsize], sigma[fontsize])
sift = cv2.xfeatures2d.SIFT_create()
kp, des = sift.detectAndCompute(img,None)
if(des is None):
des = np.array([])
data['name'] = candidate
data[os.path.join(path, 'blurred_img')] = blurred_img
data[os.path.join(path, 'descriptor')] = des
data[os.path.join(path, 'keypoints')] = serial.serialize_keypoints(kp)
data[os.path.join(path, 'length')] = image.get_text_dim(candidate, fontsize, fontfamily)[0]
return data
def create_preprocessing_data(urls: list, name: str = 'alexa') -> None:
start_time = time.time()
for url in urls:
for fontfamily in fontfamilies:
for fontsize in fontsizes:
directory = os.path.join(data_path, name, url, fontfamily, str(fontsize))
if not os.path.exists(directory):
os.makedirs(directory)
length = image.get_text_dim(url, fontsize, fontfamily)[0]
img = image.create_image(url, fontsize, fontfamily)
blurred_img = image.blur(img, kernel[fontsize], sigma[fontsize])
sift = cv2.xfeatures2d.SIFT_create()
kp, des = sift.detectAndCompute(img,None)
if(des is None):
des = np.array([])
np.save(os.path.join(data_path, name, url, fontfamily, str(fontsize), 'length'), np.array(length))
np.save(os.path.join(data_path, name, url, fontfamily, str(fontsize), 'blurred_img'), blurred_img)
np.save(os.path.join(data_path, name, url, fontfamily, str(fontsize), 'descriptor'), des)
np.save(os.path.join(data_path, name, url, fontfamily, str(fontsize), 'keypoints'), np.array(serial.serialize_keypoints(kp)))
print("--- %s seconds ---" % (time.time() - start_time))