class OpenFaceRecognition(object):
def __init__(self,
detector_model_filename,
detector_weights_filename,
detector_config_filename,
frontalize=False,
metric_distance=euclidean):
self.face_detector = OpenCVDNNDetector(detector_model_filename,
detector_weights_filename,
detector_config_filename, 300)
self.detector_model_filename = detector_model_filename
self.facial_features_estimator = FacialFeaturesEstimator(
face_3d_model_filename, embedding_model_file)
self.detector_weights_filename = detector_weights_filename
self.frontalize = frontalize
self.metric_distance = metric_distance
def extract(self, rgb_image):
face_list = self.face_detector.detect(rgb_image)
if len(face_list) == 0:
print("no image found for extraction")
return []
else:
self.facial_features_estimator.estimate(rgb_image, face_list,
self.frontalize)
name = self.facial_features_estimator.name
return face_list[0].features[name]
def predict(self, rgb_image_1, rgb_image_2):
feature1 = self.extract(rgb_image_1)
feature2 = self.extract(rgb_image_2)
return (1 -
self.metric_distance(feature1.to_array(), feature2.to_array()))
class OpenFaceRecognition(object):
def __init__(self,
detector_model_filename,
detector_weights_filename,
detector_config_filename,
face_3d_model_filename,
embedding_model_file,
shape_predictor_config_filename,
frontalize=False,
metric_distance=euclidean):
self.face_detector = OpenCVDNNDetector(detector_model_filename,
detector_weights_filename,
detector_config_filename, 300)
self.detector_model_filename = detector_model_filename
self.facial_landmarks_estimator = FacialLandmarksEstimator(
shape_predictor_config_filename)
self.face_alignement_estimator = FaceAlignementEstimator()
self.facial_features_estimator = FacialFeaturesEstimator(
face_3d_model_filename, embedding_model_file, frontalize)
#self.input_shape = input_shape
self.detector_weights_filename = detector_weights_filename
self.frontalize = frontalize
self.metric_distance = metric_distance
def extract(self, rgb_image):
face_list = self.face_detector.detect(rgb_image)
if len(face_list) == 0:
print("no image found for extraction")
return None
else:
# cv2.imshow('image',rgb_image)
# cv2.waitKey(0)
self.facial_landmarks_estimator.estimate(rgb_image, face_list)
self.facial_features_estimator.estimate(rgb_image, face_list,
self.frontalize)
name = self.facial_features_estimator.name
# rgb_image2 = self.face_alignement_estimator.align(rgb_image,face_list[0])
# cv2.imshow('image',rgb_image2)
# cv2.waitKey(0)
return face_list[0].features[name]
def predict(self, rgb_image_1, rgb_image_2):
feature1 = self.extract(rgb_image_1)
feature2 = self.extract(rgb_image_2)
return (1 -
self.metric_distance(feature1.to_array(), feature2.to_array()))
detector_config_filename, 300)
try:
os.makedirs(snapshot_directory)
except OSError as e:
if not os.path.isdir(snapshot_directory):
raise RuntimeError("{}".format(e))
snapshot_index = 0
capture = cv2.VideoCapture(0)
while True:
ok, frame = capture.read()
viz_frame = frame.copy()
if ok:
rgb_image = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
face_list = face_detector.detect(rgb_image)
if len(face_list) > 0:
# _,a,score = knn.predict(model.extract(rgb_image).to_array())
# face_list[0].confidence = score[0]
# face_list[0].label += " " + a
color = (0, 230, 0)
# if a != "alexandre":
# color = (251,0,0)
face_list[0].draw(frame, color)
k = cv2.waitKey(1) & 0xFF
if k == 32 and len(face_list) > 0:
print("Save image " + str(snapshot_index) + ".jpg !")
cv2.imwrite(snapshot_directory + str(snapshot_index) + ".jpg",
viz_frame)
snapshot_index += 1
cv2.imshow("Snapshot recorder", (255 - frame))