def test_database(self):
image = read_image("./test/data/test.jpg")
bboxes, landmarks = self.sdk.detect_faces(image)
descriptors = []
for user_id, (bbox, keypoints) in enumerate(zip(bboxes, landmarks)):
face = self.sdk.align_face(image, keypoints)
descriptor = self.sdk.get_descriptor(face)
descriptors.append(descriptor)
self.sdk.add_descriptor(descriptor, user_id)
db_path = "./test/data/test.index"
self.sdk.save_database(db_path)
self.sdk.load_database(db_path)
os.remove(db_path)
for user_id, descriptor in enumerate(descriptors):
found_ids, distances = self.sdk.find_most_similar(descriptor,
top_k=5)
assert found_ids[0] == user_id
assert np.allclose(distances[0], 1.0)
def test_detector(self):
image = read_image("./test/data/test.jpg")
bboxes, landmarks = self.sdk.detect_faces(image)
assert len(bboxes) == len(landmarks) == 5
def test_alignment(self):
image = read_image("./test/data/test.jpg")
landmarks = np.array([
[1126.5667, 306.16245],
[1245.9434, 301.26337],
[1169.3936, 383.2747],
[1142.5924, 418.49274],
[1260.0977, 413.04846],
]).T
assert landmarks.shape == (2, 5)
face = self.sdk.align_face(image, landmarks)
assert face.shape == (112, 112, 3)
save_image(face, "./test/data/alignment_result.jpg")
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--path", "-p", help="path to image", type=str)
parser.add_argument("--config", help="path to sdk config", type=str, default="face_recognition_sdk/config/config.yaml")
args = parser.parse_args()
config = read_yaml(args.config) if args.config else None
sdk = FaceRecognitionSDK(config=config)
device = "cuda" if torch.cuda.is_available() else "cpu"
img_path = Path(args.path).expanduser().resolve()
img = read_image(img_path.as_posix())
results = []
for i in range(30):
start_time = time.time()
t = sdk.recognize_faces(img)
if device == "cuda":
torch.cuda.synchronize()
time_taken = time.time() - start_time
results.append(time_taken)
results = np.asarray(results)
def test_similarity(self):
face = read_image("./test/data/alignment_result.jpg")
similarity, _, _ = self.sdk.verify_faces(face, face)
assert similarity == 1.0