test_label_file = open(os.path.join('label_data', 'test_complete_file.txt'),
'r')
for line in test_label_file.readlines():
line_input = line.split(" ")
label_dict_test[line_input[0]] = line_input[2].rstrip()
test_label_file.close()
test_labels = []
test_features = np.empty((len(files), 60))
for i, file in enumerate(files):
# Get label for file
file_name = file
# Extract lip as 60x80 image
file = os.path.join('test_data', file)
fe = FeatureExtractor.from_image(file)
fe.face_detect()
fe.landmark_detect()
fe.crop_lips()
# print("File {} lips extracted.".format(i))
# Convert lip image to 60-dimensional feature vector
if fe.lips is None:
print("File {} skipped, lips could not be found.".format(i))
continue
filters = MSA.multiscale_full(fe.lips[0].flatten('F'))
differences = filters[1:] - filters[:-1]
test_features[i] = np.sum(differences, 1)
test_labels.append(label_dict_test[file_name])
# print("File {} complete.".format(i))
