blob = cv2.dnn.blobFromImage(
img) # 'size' param resize the output to the given shape
# Load the net
net = cv2.dnn.readNet(args.model)
net.setPreferableBackend(args.backend)
net.setPreferableTarget(args.target)
# Run the net
output_names = ['loc', 'conf', 'iou']
net.setInput(blob)
loc, conf, iou = net.forward(output_names)
# Decode bboxes and landmarks
pb = PriorBox(input_shape=(w, h), output_shape=(w, h))
dets = pb.decode(np.squeeze(loc, axis=0), np.squeeze(conf, axis=0),
np.squeeze(iou, axis=0), args.conf_thresh)
# NMS
if dets.shape[0] > 0:
dets = nms(dets, args.nms_thresh)
faces = dets[:args.keep_top_k, :]
print('Detection results: {} faces found'.format(faces.shape[0]))
print(faces)
else:
print('No faces found.')
exit()
# Draw boudning boxes and landmarks on the original image
img_res = draw(cv2.cvtColor(img, cv2.COLOR_BGR2RGB), faces[:, :4],
np.reshape(faces[:, 4:14], (-1, 5, 2)), faces[:, -1])
if args.vis:
dst=None,
dsize=(input_shape),
interpolation=cv2.INTER_LINEAR)
hr, wr, _ = img_resize.shape
print('Network input size: h={}, w={}'.format(hr, wr))
blob = cv2.dnn.blobFromImage(img_resize, size=input_shape)
# run the net
output_names = ['loc', 'conf']
net.setInput(blob)
loc, conf = net.forward(output_names)
# Decode bboxes and landmarks
pb = PriorBox(input_shape=input_shape, output_shape=(w, h))
dets = pb.decode(np.squeeze(loc, axis=0), np.squeeze(conf, axis=0))
# Ignore low scores
idx = np.where(dets[:, -1] > args.conf_thresh)[0]
dets = dets[idx]
# NMS
if dets.shape[0] > 0:
dets = nms(dets, args.nms_thresh)
faces = dets[:args.keep_top_k, :]
print('Detection results: {} faces found'.format(faces.shape[0]))
print(faces)
else:
print('No faces found.')
exit()
class faceDetectorModel:
def __init__(self,
method='haarCascades',
gpu=0,
confidence=0.7,
threshold=0.3):
self.gpu = gpu
self.method = method
self.init = 0
self.detector = None
self.pb = None
self.detectorInit()
self.confidence = confidence
self.threshold = threshold
def detectorInit(self):
if self.method == 'haarCascades':
if self.gpu == 0:
self.detector = cv2.CascadeClassifier(
'faceDetect/haarcascade_frontalface_default.xml')
elif self.gpu == 1:
self.detector = cv2.cuda.CascadeClassifier_create(
'faceDetect'
'/haarcascade_frontalface_default_cuda.xml')
elif self.method == 'lbpCascades':
if self.gpu == 0:
self.detector = cv2.CascadeClassifier(
'faceDetect/lbpcascade_frontalface_improved.xml')
elif self.gpu == 1:
self.detector = cv2.cuda.CascadeClassifier_create(
'faceDetect/lbpcascade_frontalface_improved.xml')
if self.method == 'yuNet':
self.detector = cv2.dnn.readNet(
'faceDetect/YuFaceDetectNet_640.onnx')
self.detector.setPreferableBackend(cv2.dnn.DNN_BACKEND_OPENCV)
self.detector.setPreferableTarget(cv2.dnn.DNN_TARGET_CPU)
def CascadesDetector(self, frame):
if self.gpu == 1:
faces = []
gpuFrame = cv2.cuda_GpuMat()
gpuFrame.upload(frame)
gpuMat = cv2.cuda.cvtColor(gpuFrame, cv2.COLOR_BGR2GRAY)
objbuff = self.detector.detectMultiScale(gpuMat)
facess = objbuff.download()
if facess is None:
facess = ()
np.array(facess)
for multipleFace in facess:
for face in multipleFace:
faces.append(face)
return faces
elif self.gpu == 0:
grayFrame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
faces = self.detector.detectMultiScale(grayFrame,
scaleFactor=1.2,
minNeighbors=5,
minSize=(20, 20))
return faces
def yuNetDetection(self, frame):
if self.init == 0:
frameWidth, frameHeight = frame.shape[:2]
self.pb = PriorBox(input_shape=(640, 480),
output_shape=(frameHeight, frameWidth))
self.init = 1
blob = cv2.dnn.blobFromImage(frame, size=(640, 480))
outputNames = ['loc', 'conf', 'iou']
self.detector.setInput(blob)
loc, conf, iou = self.detector.forward(outputNames)
dets = self.pb.decode(np.squeeze(loc, axis=0), np.squeeze(conf,
axis=0),
np.squeeze(iou, axis=0))
idx = np.where(dets[:, -1] > self.confidence)[0]
dets = dets[idx]
if dets.shape[0]:
facess = nms(dets, self.threshold)
else:
facess = ()
return facess
faces = np.array(facess[:, :4])
faces = faces.astype(np.int)
faceStartXY = faces[:, :2]
faceEndXY = faces[:, 2:4]
faceWH = faceEndXY - faceStartXY
faces = np.hstack((faceStartXY, faceWH))
# scores = facess[:, -1]
return faces
def predict(self, frame, painted=1):
frameNew = frame.copy()
faces = ()
if self.method == 'haarCascades' or self.method == 'lbpCascades':
faces = self.CascadesDetector(frameNew)
elif self.method == 'yuNet':
faces = self.yuNetDetection(frameNew)
if painted:
for (x, y, w, h) in faces:
cv2.rectangle(frameNew, (x, y), (x + w, y + h), (0, 0, 255))
return frameNew, faces