def main():
fps = 0.0
minsize = 20
threshold = [0.06, 0.6, 0.6]
factor = 0.709
Pnet, Rnet, Onet = detector.initFaceDetector()
video_capture = cv2.VideoCapture(0)
while True:
t1 = time.time()
_, img = video_capture.read()
cv2.imshow("origin", img)
cv2.waitKey(10)
img_matlab = img.copy()
tmp = img_matlab[:, :, 2].copy()
img_matlab[:, :, 2] = img_matlab[:, :, 0] #### BGR CHW
img_matlab[:, :, 0] = tmp
boundingboxes = detector.detect_face(img_matlab, minsize, Pnet, Rnet,
Onet, threshold, False, factor)
detector.drawBoxes(img, boundingboxes)
cv2.imshow("Pnet", img)
fps = (fps + (1. / (time.time() - t1))) / 2
print("fps= %f" % (fps))
# Press Q to stop!
if cv2.waitKey(1) & 0xFF == ord('q'):
break
video_capture.release()
cv2.destroyAllWindows()
def main(use_gpu=False):
frame_root = 'test/data/3.0'
res_root = 'test/res/3.0/face_attribute'
if not os.path.exists(res_root):
os.makedirs(res_root)
for vid in tqdm(sorted(os.listdir(frame_root))):
video_frame_root = os.path.join(frame_root, vid)
fids, images = get_input(video_frame_root)
results = []
for i, image in enumerate(images):
face_box = detect_face(image, det_net)
if face_box is None:
res = ['NaN'] * len(attr_names)
else:
face_image = crop_face(face_box, image, 2.4)
attr_probs = recognize_attributes(face_image, rec_net, use_gpu)
res = [attr_probs[i] for i in range(len(attr_probs))]
res = ['%.4f' % scr for scr in res]
results.append(res)
res_path = os.path.join(res_root, vid + '.txt')
with open(res_path, 'w') as f:
for i in range(len(fids)):
line = '%s %s\n' % (fids[i].split('.')[0], ' '.join(results[i]))
f.writelines(line)
def detect(im):
bounding_boxes, _ = detector.detect_face(im, minsize, pnet, rnet, onet, threshold, factor)
if len(bounding_boxes) < 1:
return None
det = np.squeeze(bounding_boxes[0, 0:4])
bb = np.zeros(4, dtype=np.int32)
bb[0] = np.maximum(det[0] - margin / 2, 0)
bb[1] = np.maximum(det[1] - margin / 2, 0)
bb[2] = np.minimum(det[2] + margin / 2, im.shape[1])
bb[3] = np.minimum(det[3] + margin / 2, im.shape[0])
cropped = im[bb[1]:bb[3], bb[0]:bb[2], :]
im = Image.fromarray(cropped, 'RGB')
im = im.resize((160, 160))
im = ops.normalize_image(im)
return im, bb
def main(video_list_path, res_root, use_gpu=False):
global logger
with open(video_list_path) as f:
video_list = [line.strip().split()[1] for line in f.readlines()]
if not os.path.exists(res_root):
os.makedirs(res_root)
for frame_dir in tqdm(video_list):
video_frame_root = frame_dir
fids, images = get_input(video_frame_root)
vid = frame_dir.split('/')[-1]
res_path = os.path.join(res_root, vid + '.txt')
if os.path.exists(res_path):
with open(res_path, 'r') as f:
lines = f.readlines()
if len(lines) == len(fids):
continue
results = []
for i, image in enumerate(images):
face_box = detect_face(image, det_net)
if face_box is None:
res = ['NaN'] * len(attr_names)
else:
face_image = crop_face(face_box, image, 2.4)
attr_probs = recognize_attributes(face_image, rec_net, use_gpu)
res = [attr_probs[i] for i in range(len(attr_probs))]
res = ['%.4f' % scr for scr in res]
results.append(res)
with open(res_path, 'w') as f:
for i in range(len(fids)):
line = '%s %s\n' % (fids[i].split('.')[0], ' '.join(
results[i]))
f.writelines(line)
logger.write('%s\n' % vid)
logger.flush()
logger.write('%s done!' % video_list_path)
logger.flush()
logger.close()
def mark(requests):
"""
:param requests: Image
:return: Cropped face
"""
im = ops.read_image(requests)
bounding_boxes, _ = detector.detect_face(im, minsize, pnet, rnet, onet, threshold, factor)
if len(bounding_boxes) < 1:
return HttpResponse('no face detected.')
det = np.squeeze(bounding_boxes[0, 0:4])
bb = np.zeros(4, dtype=np.int32)
bb[0] = np.maximum(det[0] - margin / 2, 0)
bb[1] = np.maximum(det[1] - margin / 2, 0)
bb[2] = np.minimum(det[2] + margin / 2, im.shape[1])
bb[3] = np.minimum(det[3] + margin / 2, im.shape[0])
return HttpResponse(' '.join(map(lambda i: str(i), list(bb))))
def main(body):
logger, log_stream = get_logger()
image = None
success = False
data = None
try:
file = process_body(body)
except Exception:
file = None
logger.error('bad request body!', exc_info=True, stack_info=True)
if file is not None:
try:
image = decode_img_from_b64(file)
except Exception:
logger.error('bad image!', exc_info=True, stack_info=True)
if image is not None:
try:
face_box = detect_face(image, det_net)
if face_box is None:
logger.error('no face detected!')
else:
face_image = crop_face(face_box, image, 2.4)
attr_probs = recognize_attributes(face_image, rec_net)
res = ['%.4f' % prob for prob in attr_probs]
data = ' '.join(res)
success = True
except Exception:
logger.error('model runtime error!',
exc_info=True,
stack_info=True)
res = {'data': data, 'success': success, 'log': log_stream.getvalue()}
return res
# for debugging
# statusLight = gpiozero.LED(4)
# statusLightSwitch = 1
while True:
# indicate if the program is still running
# statusLightSwitch *= -1
# if statusLightSwitch > 0:
# statusLight.on()
# else:
# statusLight.off()
# default for the program is to detect every 6 seconds
time.sleep(6) # call detector every 6 seconds
s = detector.detect_face() # detection results: true for face detected
timeline.pop(0)
if s:
timeline.append(1)
else:
timeline.append(0)
# more calculation, but simplifies the design logic
sit = sum(timeline)
if sit >= 500:
led.on()
else:
led.off()
# if leave for 5 minutes, reset timeline
if sum(timeline[-50:]) == 0: # -50