def recognize_iris(frame, detector, estimator, estimator2, out_frame=None):
_, img128, scale, pad = iut.resize_pad(frame[:, :, ::-1])
input_data = img128.astype('float32') / 127.5 - 1.0
input_data = np.expand_dims(np.moveaxis(input_data, -1, 0), 0)
# inference
# Face detection
preds = detector.predict([input_data])
detections = iut.detector_postprocess(
preds, anchor_path="../../face_recognition/facemesh/anchors.npy")
# display bbox
if args.bbox:
detections2 = iut.denormalize_detections(detections[0].copy(), scale,
pad)
display_hand_box(out_frame, detections2)
# Face landmark estimation
if detections[0].size != 0:
imgs, affines, box = iut.estimator_preprocess(frame[:, :, ::-1],
detections, scale, pad)
landmarks = np.zeros((imgs.shape[0], 1404))
normalized_landmarks = np.zeros((imgs.shape[0], 468, 3))
confidences = np.zeros((imgs.shape[0], 1))
for i in range(imgs.shape[0]):
landmark, confidences[i, :] = estimator.predict(
[imgs[i:i + 1, :, :, :]])
normalized_landmark = landmark / 192.0
normalized_landmarks[i, :, :] = fut.denormalize_landmarks(
normalized_landmark, affines)
landmarks[i, :] = landmark
#Added
for i in range(len(normalized_landmark)):
landmark, confidence = normalized_landmarks[i], confidences[i]
draw_landmarks_face(out_frame, landmark[:, :2], size=1)
# Iris landmark estimation
imgs2, origins = iut.iris_preprocess(imgs, landmarks)
eyes = np.zeros((imgs2.shape[0], 213))
iris = np.zeros((imgs2.shape[0], 15))
for i in range(imgs2.shape[0]):
eyes[i, :], iris[i, :] = estimator2.predict(
[imgs2[i:i + 1, :, :, :]])
eyes, iris = iut.iris_postprocess(eyes, iris, origins, affines)
for i in range(len(eyes)):
draw_eye_iris(out_frame,
eyes[i, :, :16, :2],
iris[i, :, :, :2],
size=2)
def recognize_from_image():
# prepare input data
src_img = cv2.imread(args.input)
_, img128, scale, pad = fut.resize_pad(src_img[:,:,::-1])
input_data = img128.astype('float32') / 127.5 - 1.0
input_data = np.expand_dims(np.moveaxis(input_data, -1, 0), 0)
# net initialize
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
detector = ailia.Net(DETECTION_MODEL_PATH, DETECTION_WEIGHT_PATH, env_id=env_id)
estimator = ailia.Net(LANDMARK_MODEL_PATH, LANDMARK_WEIGHT_PATH, env_id=env_id)
# inference
print('Start inference...')
if args.benchmark:
print('BENCHMARK mode')
for _ in range(5):
start = int(round(time.time() * 1000))
# Face detection
preds = detector.predict([input_data])
detections = fut.detector_postprocess(preds)
# Face landmark estimation
if detections[0].size != 0:
imgs, affines, box = fut.estimator_preprocess(src_img[:,:,::-1], detections, scale, pad)
draw_roi(src_img, box)
estimator.set_input_shape(imgs.shape)
landmarks, confidences = estimator.predict([imgs])
normalized_landmarks = landmarks / 192.0
# postprocessing
landmarks = fut.denormalize_landmarks(normalized_landmarks, affines)
for i in range(len(landmarks)):
landmark, confidence = landmarks[i], confidences[i]
# if confidence > 0: # Can be > 1, no idea what it represents
draw_landmarks(src_img, landmark[:,:2], size=1)
end = int(round(time.time() * 1000))
print(f'\tailia processing time {end - start} ms')
else:
# Face detection
preds = detector.predict([input_data])
detections = fut.detector_postprocess(preds)
# Face landmark estimation
if detections[0].size != 0:
imgs, affines, box = fut.estimator_preprocess(src_img[:,:,::-1], detections, scale, pad)
draw_roi(src_img, box)
estimator.set_input_shape(imgs.shape)
landmarks, confidences = estimator.predict([imgs])
normalized_landmarks = landmarks / 192.0
# postprocessing
landmarks = fut.denormalize_landmarks(normalized_landmarks, affines)
for i in range(len(landmarks)):
landmark, confidence = landmarks[i], confidences[i]
# if confidence > 0: # Can be > 1, no idea what it represents
draw_landmarks(src_img, landmark[:,:2], size=1)
cv2.imwrite(args.savepath, src_img)
print('Script finished successfully.')
def recognize_from_video():
# net initialize
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
detector = ailia.Net(DETECTION_MODEL_PATH, DETECTION_WEIGHT_PATH, env_id=env_id)
estimator = ailia.Net(LANDMARK_MODEL_PATH, LANDMARK_WEIGHT_PATH, env_id=env_id)
capture = get_capture(args.video)
# create video writer if savepath is specified as video format
if args.savepath != SAVE_IMAGE_PATH:
f_h = int(capture.get(cv2.CAP_PROP_FRAME_HEIGHT))
f_w = int(capture.get(cv2.CAP_PROP_FRAME_WIDTH))
save_h, save_w = webcamera_utils.calc_adjust_fsize(
f_h, f_w, IMAGE_HEIGHT, IMAGE_WIDTH
)
writer = webcamera_utils.get_writer(args.savepath, save_h, save_w)
else:
writer = None
while(True):
ret, frame = capture.read()
if (cv2.waitKey(1) & 0xFF == ord('q')) or not ret:
break
frame = np.ascontiguousarray(frame[:,::-1,:])
_, img128, scale, pad = fut.resize_pad(frame[:,:,::-1])
input_data = img128.astype('float32') / 127.5 - 1.0
input_data = np.expand_dims(np.moveaxis(input_data, -1, 0), 0)
# inference
# Face detection
preds = detector.predict([input_data])
detections = fut.detector_postprocess(preds)
# Face landmark estimation
if detections[0].size != 0:
imgs, affines, box = fut.estimator_preprocess(frame[:,:,::-1], detections, scale, pad)
draw_roi(frame, box)
dynamic_input_shape = False
if dynamic_input_shape:
estimator.set_input_shape(imgs.shape)
landmarks, confidences = estimator.predict([imgs])
normalized_landmarks = landmarks / 192.0
landmarks = fut.denormalize_landmarks(normalized_landmarks, affines)
else:
landmarks = np.zeros((imgs.shape[0], 468, 3))
confidences = np.zeros((imgs.shape[0], 1))
for i in range(imgs.shape[0]):
landmark, confidences[i,:] = estimator.predict([imgs[i:i+1,:,:,:]])
normalized_landmark = landmark / 192.0
landmarks[i,:,:] = fut.denormalize_landmarks(normalized_landmark, affines)
for i in range(len(landmarks)):
landmark, confidence = landmarks[i], confidences[i]
# if confidence > 0: # Can be > 1, no idea what it represents
draw_landmarks(frame, landmark[:,:2], size=1)
cv2.imshow('frame', frame)
# save results
if writer is not None:
writer.write(frame)
capture.release()
cv2.destroyAllWindows()
print('Script finished successfully.')
pass