def gen_train(path,label):
res = ""
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(sys.argv[2])
for f in glob.glob(os.path.join(path, "*.jpg")):
print("Processing file: {}".format(f))
img = io.imread(f)
# Ask the detector to find the bounding boxes of each face. The 1 in the
# second argument indicates that we should upsample the image 1 time. This
# will make everything bigger and allow us to detect more faces.
dets = detector(img, 1)
print("Number of faces detected: {}".format(len(dets)))
for k, d in enumerate(dets):
# print("Detection {}: Left: {} Top: {} Right: {} Bottom: {}".format(
# k, d.left(), d.top(), d.right(), d.bottom()))
# Get the landmarks/parts for the face in box d.
shape = predictor(img, d)
# print("Part 0: {}, Part 1: {} ...".format(shape.part(0),
# shape.part(1)))
val = feature_calc(shape)
res += str(label) + ' '
for x in range(1,len(val)+1):
res += str(x) + ':' + str(val[x-1])
if x != len(val):
res += ' '
else:
res += '\n'
with open(str(label)+'.csv','w') as fout:
fout.write(res)
def calcPoints(path) :
predictor_path = '/Users/vovanmozg/Downloads/bigdata/shape_predictor_68_face_landmarks.dat'
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(predictor_path)
print("Processing file: {}".format(path))
img = io.imread(path)
# Ask the detector to find the bounding boxes of each face. The 1 in the
# second argument indicates that we should upsample the image 1 time. This
# will make everything bigger and allow us to detect more faces.
dets = detector(img, 1)
if len(dets) != 1 :
return False;
print("Number of faces detected: {}".format(len(dets)))
points = [];
for k, d in enumerate(dets):
#print("Detection {}: Left: {} Top: {} Right: {} Bottom: {}".format(
# k, d.left(), d.top(), d.right(), d.bottom()))
# Get the landmarks/parts for the face in box d.
shape = predictor(img, d)
#print(numpy.matrix([[p.x, p.y] for p in shape.parts()]))
for p in shape.parts():
points.append((p.x, p.y))
return points
def load_model(self):
self.predictor = dlib.shape_predictor(self.predictor_path)
self.detector = dlib.get_frontal_face_detector()
self.mean3DShape, self.blendshapes, self.mesh, self.idxs3D, self.idxs2D = utils.load3DFaceModel(self.FaceModel_file)
self.projectionModel = models.OrthographicProjectionBlendshapes(self.blendshapes.shape[0])
self.renderer = FaceRendering.FaceRenderer(width=self.width, height=self.height)
self.projectionModel = models.OrthographicProjectionBlendshapes(self.blendshapes.shape[0])
def process_frames_face(self, frames):
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(self.face_predictor_path)
mouth_frames = self.get_frames_mouth(detector, predictor, frames)
self.face = np.array(frames)
self.mouth = np.array(mouth_frames)
self.set_data(mouth_frames)
def __init__(self, facePredictor, padding=0.2):
"""
"""
assert facePredictor is not None
self.detector = dlib.get_frontal_face_detector()
self.predictor = dlib.shape_predictor(facePredictor)
self.padding = padding
# add padding
new_template =[]
for item in TEMPLATE:
new_item = ((item[0]+self.padding)/(2*self.padding+1),(item[1]+self.padding)/(2*self.padding+1))
new_template.append(new_item)
self.new_template = np.float32(new_template)
# this shape is reference from dlib implement.
self.mean_shape_x = [0.000213256, 0.0752622, 0.18113, 0.29077, 0.393397, 0.586856, 0.689483, 0.799124,
0.904991, 0.98004, 0.490127, 0.490127, 0.490127, 0.490127, 0.36688, 0.426036,
0.490127, 0.554217, 0.613373, 0.121737, 0.187122, 0.265825, 0.334606, 0.260918,
0.182743, 0.645647, 0.714428, 0.793132, 0.858516, 0.79751, 0.719335, 0.254149,
0.340985, 0.428858, 0.490127, 0.551395, 0.639268, 0.726104, 0.642159, 0.556721,
0.490127, 0.423532, 0.338094, 0.290379, 0.428096, 0.490127, 0.552157, 0.689874,
0.553364, 0.490127, 0.42689]#17-67
self.mean_shape_y = [0.106454, 0.038915, 0.0187482, 0.0344891, 0.0773906, 0.0773906, 0.0344891,
0.0187482, 0.038915, 0.106454, 0.203352, 0.307009, 0.409805, 0.515625, 0.587326,
0.609345, 0.628106, 0.609345, 0.587326, 0.216423, 0.178758, 0.179852, 0.231733,
0.245099, 0.244077, 0.231733, 0.179852, 0.178758, 0.216423, 0.244077, 0.245099,
0.780233, 0.745405, 0.727388, 0.742578, 0.727388, 0.745405, 0.780233, 0.864805,
0.902192, 0.909281, 0.902192, 0.864805, 0.784792, 0.778746, 0.785343, 0.778746,
0.784792, 0.824182, 0.831803, 0.824182] #17-67
def __init__(self):
self.cap = cv2.VideoCapture(0)
self.fourcc = cv2.cv.CV_FOURCC(*'XVID')
self.out = cv2.VideoWriter('data/' + subject + '/' + video +'/output.avi', -1, 9, (width, height))
self.detector = dlib.get_frontal_face_detector()
self.predictor = dlib.shape_predictor("shape_predictor_68_face_landmarks.dat")
# Imaging device - must be a connected camera (not an ip camera or mjpeg
# stream)
self.camera = Camera(camera=0) # first camera by default
self.w, self.h = 0, 0
self.pressed = 0
# Containerized analysis of recieved image frames (an openMDAO assembly)
# is defined next.
# This assembly is designed to handle all image & signal analysis,
# such as face detection, forehead isolation, time series collection,
# heart-beat detection, etc.
# Basically, everything that isn't communication
# to the camera device or part of the GUI
self.processor = findFaceGetPulse(bpm_limits=[50, 160],
data_spike_limit=2500.,
face_detector_smoothness=10.)
# Init parameters for the cardiac data plot
self.bpm_plot = False
self.plot_title = "Cardiac info - raw signal, filtered signal, and PSD"
# Maps keystrokes to specified methods
# (A GUI window must have focus for these to work)
self.key_controls = {"s": self.toggle_search,
"d": self.toggle_display_plot,
"f": self.write_csv}
def __init__(self, landmarks=None, openface=None, size=96, torch="th"):
"""Face detection
Parameters
----------
landmarks : str
Path to dlib's 68 facial landmarks predictor model.
size : int
Size of the normalized face thumbnail.
openface : str
Path to openface FaceNet model.
"""
super(Face, self).__init__()
# face detection
self._face_detector = dlib.get_frontal_face_detector()
# landmark detection
if landmarks is not None:
self._landmarks_detector = dlib.shape_predictor(landmarks)
# normalization
self.size = size
self._landmarks = MINMAX_TEMPLATE * self.size
if openface is not None:
self._net = TorchWrap(torch=torch, model=openface, size=self.size, cuda=False)
def feature_extraction(img):
dist = [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]
dist = np.asarray(dist)
#Loading the facial land mark pointer predictor
predictor_path = '/home/qburst/Desktop/Emotion Detection Final/ED_dist/Necessaries/shape_predictor_68_face_landmarks.dat'
#Creating objects for frontal face detector and landmark point predictor
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(predictor_path)
#'dets' has all the faces detected represented as rectangles, their co-ordinates are obtainable by using left(), top(),
#right(), bottom(). 'dets' is iterable
dets = detector(img, 1)
for rect in dets:
#'shape' has the 68 facial land mark points. Each of which can be accessed using shape.part().co-ordinate
#eg: shape.part(1).x, shape.part(1).y for (x, y) of point 1. 'shape' is not iterable.
shape = predictor(img, rect)
points = _shape_to_np(shape)
dist_as_np = np.asarray(distance(points))
dist = np.vstack((dist, dist_as_np))
dist = np.delete(dist, 0, 0)
return len(dets), dist, dets
def __init__(self, predictor, detector=None):
if detector is None:
self.detector = dlib.get_frontal_face_detector()
else:
self.detector = detector
self.predictor = predictor
self.reset()
def __init__(self):
self._detector = dlib.get_frontal_face_detector()
pose_predictor_path = '../models/dlib/shape_predictor_68_face_landmarks.dat'
self._predictor = dlib.shape_predictor(pose_predictor_path)
if not self.SERVER_NO_GUI_MODE:
self._win = dlib.image_window()
self._lock = RLock()
def main():
detector = dlib.get_frontal_face_detector()
try:
src=sys.argv[1]
except:
src=0
cap=videoCap(src)
while True:
ret,frame=cap.read()
t=clock()
rects=detector(frame,1)
vis=frame.copy()
draw_dlib_rects(vis,rects,(0,255,0))
dt=clock()-t
draw_str(vis,(20,20),'time: %.1f ms' % (dt*1000))
cv2.imshow('facedetect',vis)
if 0xFF & cv2.waitKey(5) == 27:
break
cv2.destroyAllWindows()
def __init__(self): self.PREDICTOR_PATH = "shape_predictor_68_face_landmarks.dat" MOUTH_POINTS = list(range(48, 61)) self.OVERLAY_POINTS = [MOUTH_POINTS] self.detector = dlib.get_frontal_face_detector() self.predictor = dlib.shape_predictor(self.PREDICTOR_PATH)
def camera_fd():
import cv2
cam = cv2.VideoCapture(-1)
win = dlib.image_window()
detector = detector = dlib.get_frontal_face_detector()
def detect():
s, img = cam.read()
if not s:
return
img2 = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
dets = detector(img2)
num = len(dets)
if num > 0:
print("found!")
win.clear_overlay()
win.set_image(img2)
win.add_overlay(dets)
try:
while True:
detect()
except KeyboardInterrupt as e:
print("exiting")
cam.release()
cv2.destroyAllWindows()
def get_landmarks(line_arr):
predictor_path = 'shape_predictor_68_face_landmarks.dat' # http://sourceforge.net/projects/dclib/files/dlib/v18.10/shape_predictor_68_face_landmarks.dat.bz2
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(predictor_path)
lmarks = []
bboxes = []
for i,line in enumerate(line_arr):
print('%d/%d'%(i,len(line_arr)))
img = io.imread(line)
dets = detector(img, 0)
if len(dets) == 0:
rect = dlib.rectangle(0,0,img.shape[0], img.shape[1])
else:
rect = dets[0]
shape = predictor(img, rect)
xy = _shape_to_np(shape)
lmarks.append(xy)
bboxes.append(rect)
lmarks = np.vstack(lmarks)
bboxes = np.asarray(bboxes)
return lmarks,bboxes
def initialize(detector, scale_to=2048):
global dlib_detectors
global keras_model
global is_initialized
if not is_initialized:
dlib_cnn_face_detector_path = os.path.join(os.path.dirname(__file__), "mmod_human_face_detector.dat")
if not os.path.exists(dlib_cnn_face_detector_path):
raise Exception ("Error: Unable to find %s, reinstall the lib !" % (dlib_cnn_face_detector_path) )
if detector == 'cnn' or detector == "all":
dlib_cnn_face_detector = dlib.cnn_face_detection_model_v1(dlib_cnn_face_detector_path)
#DLIB and TF competiting for VRAM, so dlib must do first allocation to prevent OOM error
dlib_cnn_face_detector ( np.zeros ( (scale_to, scale_to, 3), dtype=np.uint8), 0 )
dlib_detectors.append(dlib_cnn_face_detector)
if detector == "hog" or detector == "all":
dlib_face_detector = dlib.get_frontal_face_detector()
dlib_face_detector ( np.zeros ( (scale_to, scale_to, 3), dtype=np.uint8), 0 )
dlib_detectors.append(dlib_face_detector)
keras_model_path = os.path.join( os.path.dirname(__file__) , "2DFAN-4.h5" )
if not os.path.exists(keras_model_path):
print ("Error: Unable to find %s, reinstall the lib !" % (keras_model_path) )
else:
print ("Info: initializing keras model...")
keras_model = keras.models.load_model (keras_model_path, custom_objects={'TorchBatchNorm2D': TorchBatchNorm2D} )
is_initialized = True
def findFace(img):
detector = dlib.get_frontal_face_detector()
orgImg = img
gray = orgGray = cv2.cvtColor(orgImg, cv2.COLOR_BGR2GRAY)
angles = (0, 5, -5, 10, -10, 15, -15, 30, -30, 45, -45, 70, -70, 90, -90)
for i in range(0, len(angles)):
dets = detector(gray, 1)
if len(dets)==1:
break
nextIndex = i+1
if nextIndex<len(angles):
#cv2.imshow("Faces found", gray)
#cv2.waitKey(0)
gray = rotateImage(orgGray, angles[nextIndex])
if len(dets)==1:
img = rotateImage(orgImg, angles[i])
for k, d in enumerate(dets):
#cv2.rectangle(img, (d.left(), d.top()), (d.right(), d.bottom()), (0, 255, 0), 2)
img = img[max(d.top(),0):d.bottom(),max(d.left(),0):d.right()]
img = cv2.resize(img, (128,128))
return img
#cv2.imwrite(outputDir + imgName, img)
#cv2.imshow("Faces found", img)
#cv2.waitKey(0)
return None
def get_landmarks(argv):
line = argv[1]
left = argv[2]
top = argv[3]
right = argv[2]+argv[4]
bottom = argv[3]+argv[5]
rect = dlib.rectangle(int(left),int(top),int(right), int(bottom))
predictor_path = 'shape_predictor_68_face_landmarks.dat' # http://sourceforge.net/projects/dclib/files/dlib/v18.10/shape_predictor_68_face_landmarks.dat.bz2
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(predictor_path)
lmarks = []
bboxes = []
#for i,line in enumerate(line_arr):
# print('%d/%d'%(i,len(line_arr)))
img = io.imread(line)
dets = detector(img, 0)
if len(dets) == 0:
rect = dlib.rectangle(0,0,img.shape[0], img.shape[1])
else:
rect = dets[0]
shape = predictor(img, rect)
xy = _shape_to_np(shape)
np.savetxt('landmarks.txt', xy)
lmarks.append(xy)
bboxes.append(rect)
lmarks = np.vstack(lmarks)
bboxes = np.asarray(bboxes)
print lmarks
return lmarks
def loadData(imagepath):
img = io.imread(imagepath)
detector = dlib.get_frontal_face_detector()
grayimg = cv2.cvtColor( img, cv2.COLOR_RGB2GRAY )
start_time = time.time()
dets = detector(grayimg,1)
print("--- %s seconds sdssd---" % (time.time() - start_time))
start_time = time.time()
imgWithRect = cv2.imread(imagepath)
ListOfFaces = []
CoordinatesList = []
for rectangle in dets:
left = rectangle.left()
top = rectangle.top()
right = rectangle.right()
bottom = rectangle.bottom()
if top<0:
offset = 0-top
top = 0
bottom = bottom+offset
if left<0:
offset = 0 -left
left =0
right = right+offset
cv2.rectangle(imgWithRect,(left,top),(right,bottom),(255,255,255),2)
crop = grayimg[top:bottom, left:right]
crop2 = cv2.resize(crop, (48, 48))
data = imagePreprocessing(crop2)
ListOfFaces.append(data)
CoordinatesList.append((left,top,right,bottom))
cv2.imwrite(imagepath,imgWithRect)
print("--- %s seconds ---" % (time.time() - start_time))
return ListOfFaces, CoordinatesList
def get_training_data(image_path, labels_path):
landmarks_predictor = face_landmarks.faceLandmarks("shape_predictor_68_face_landmarks.dat")
detector = dlib.get_frontal_face_detector()
features = []
labels = []
for dirpath, dirnames, filenames in os.walk(image_path):
files = [name for name in glob.glob(dirpath+"/*.png")]
if(len(files) != 0): #if images exist
file_path = max(files)
pure_name = os.path.splitext(os.path.basename(file_path))[0]
emotion = glob.glob(labels_path+"/*/*/"+pure_name+"_emotion.txt")
if(len(emotion) > 0): #data about emotion exist
in_file = open(emotion[0],"r")
label = int(float(in_file.read()))
in_file.close()
labels.append(label)
img = cv2.imread(file_path,cv2.IMREAD_GRAYSCALE)
dets = detector(img, 1)
#in base case will be iterate only one time (every image in CK+ dataset have one face)
for k, rect in enumerate(dets):
landmarks = landmarks_predictor.get_landmarks(img, rect)
features.append(landmarks)
return features, labels
def __init__(self):
self.bridge = CvBridge()
self.imgDim = 96
self.align = openface.AlignDlib(DLIB_FACEPREDICTOR)
self.face_pose_predictor = dlib.shape_predictor(DLIB_FACEPREDICTOR)
self.net = openface.TorchNeuralNet(NETWORK_MODEL, self.imgDim)
self.landmarkIndices = openface.AlignDlib.OUTER_EYES_AND_NOSE
self.face_detector = dlib.get_frontal_face_detector()
self.count = 0
self.face_count = 0 # Cumulative total faces in training.
self.max_face_count = 10
self.train = False
self.enable = True
self.data_root = os.path.join(CWD, 'faces')
self.train_dir = os.path.join(self.data_root, 'training-images')
self.aligned_dir = os.path.join(self.data_root, 'aligned-images')
self.classifier_dir = CLASSIFIER_DIR
self.clf, self.le = None, None
self.load_classifier(os.path.join(self.classifier_dir, 'classifier.pkl'))
self.node_name = rospy.get_name()
self.multi_faces = False
self.threshold = 0.5
self.detected_faces = deque(maxlen=10)
self.training_job = None
self.stop_training = threading.Event()
self.faces = []
self.event_pub = rospy.Publisher(
'face_training_event', String, latch=True, queue_size=1)
self.faces_pub = rospy.Publisher(
'~faces', Faces, latch=True, queue_size=1)
self.imgpub = rospy.Publisher(
'~image', Image, latch=True, queue_size=1)
self._lock = threading.RLock()
self.colors = [ (255, 0, 0), (0, 255, 0), (0, 0, 255),
(255, 255, 0), (255, 0, 255), (0, 255, 255) ]
def predictVideo(uvcID):
"""
uvcID: video camera ID
"""
detector = dlib.get_frontal_face_detector()
posePredictor = facePose.FacePosePredictor()
cap = cv2.VideoCapture(uvcID)
cv2.namedWindow("headPose and landmark q:quit", cv2.WINDOW_NORMAL)
while True:
ok, colorImage = cap.read()
if not ok:
continue
numUpSampling = 0
dets, _, _ = detector.run(colorImage, numUpSampling)
bboxs = facePose.dets2xxyys(dets)
predictpoints, landmarks, predictpose = posePredictor.predict(colorImage, bboxs)
show_image(colorImage, landmarks, bboxs, predictpose)
k = cv2.waitKey(10) & 0xff
if k == ord('q') or k == 27:
break
def detect_gaze_direction(video_capture,predictor):
cam = cv2.VideoCapture(video_capture)
cam.set(3,640)
cam.set(4,480)
video_capture = cam
detector = dlib.get_frontal_face_detector()
while True:
# Capture frame-by-frame
ret, frame = video_capture.read()
if ret:
frame_color = frame
frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
dets = detector(frame, 1)
for k, d in enumerate(dets):
# Get the landmarks/parts for the face in box d.
shape = predictor(frame, d)
# print(type(shape.part(1).x))
cv2.circle(frame_color,(shape.part(36).x,shape.part(36).y),2,(0,0,255))
cv2.circle(frame_color,(shape.part(39).x,shape.part(39).y),2,(0,0,255))
cv2.circle(frame_color,(shape.part(42).x,shape.part(42).y),2,(0,0,255))
cv2.circle(frame_color,(shape.part(45).x,shape.part(45).y),2,(0,0,255))
x1 = shape.part(36).x
y1 = shape.part(37).y-2
x2 = shape.part(39).x
y2 = shape.part(40).y+2
split = frame[y1:y2,x1:x2]
split = process_eye(split)
split = filter_eye(split)
centre = cross_spread(split)
frame[y1:y2,x1:x2]=split
y1 = y1+2
y2 = y2-2
centre[1]=centre[1]-2
# cv2.rectangle(frame_color,(x1,y1), (x2,y2), (0, 0, 255), 1)
# cv2.circle(frame_color,(x1+centre[0],y1+centre[1]),2,(0,0,255))
cv2.line(frame_color,(x1+centre[0],y1+centre[1]), (int((3*x1+4*centre[0]-x2)/2),int((3*y1+4*centre[1]-y2)/2)),(255,0,0))
x1 = shape.part(42).x
y1 = shape.part(43).y-2
x2 = shape.part(45).x
y2 = shape.part(46).y+2
split = frame[y1:y2,x1:x2]
split = process_eye(split)
split = filter_eye(split)
centre = cross_spread(split)
frame[y1:y2,x1:x2]=split
y1 = y1+2
y2 = y2-2
centre[1]=centre[1]-2
# cv2.rectangle(frame_color,(x1,y1), (x2,y2), (0, 0, 255), 1)
# cv2.circle(frame_color,(x1+centre[0],y1+centre[1]),2,(0,0,255))
cv2.line(frame_color,(x1+centre[0],y1+centre[1]), (int((3*x1+4*centre[0]-x2)/2),int((3*y1+4*centre[1]-y2)/2)),(255,0,0))
# Display the resulting frame
cv2.imshow('Video', frame_color)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
# Release video capture
video_capture.release()
cv2.destroyAllWindows()
def __init__(self, image,
with_saliency=True,
with_face_detection=True,
with_text_detection=True):
self.image = image
self.with_saliency = with_saliency
self.with_face_detection = with_face_detection
self.with_text_detection = with_text_detection
if with_face_detection:
self.haar_profile = options.haar_profile
self.profile_face_cascade = cv2.CascadeClassifier()
self.profile_face_cascade.load(self.haar_profile)
if SmartCrop._dlib_face_detector is None:
SmartCrop._dlib_face_detector = \
dlib.get_frontal_face_detector()
self.dlib_face_detector = SmartCrop._dlib_face_detector
self.haar_params = {'minNeighbors': 8,
'minSize': (100, 100),
'scaleFactor': 1.1}
self._faces = None
if with_text_detection:
self.text_classifier1 = options.text_classifier1
self.text_classifier2 = options.text_classifier2
self._text_boxes = None
self._bottom_percent = 0.7
if with_saliency:
self._saliency_map = None
self._saliency_threshold = options.saliency_threshold
def detectFace(img):
offset = 20
detector = dlib.get_frontal_face_detector()
dets = detector(img,1)
face_image = []
for d in dets:
face_image.append(img[d.top()-offset:d.bottom()+(offset*1.5), d.left()-offset:d.right()+(offset*1.5)])
return face_image, dets
def getVisualFetB():
'''
Second attempt at extracting features
Simply extract the face from each frame, followed by extracting details of the landmarks
The generated file for each video contains a numpy array of the vectors (Of facial landmarks)
'''
# fileName = '../training/training_gt.csv'
# trueMap = getTruthVal(fileName)
print 'Started extracting features B'
videoPath = '../training/download_train-val/trainFiles/'
vidNames = os.listdir(videoPath)
vidNames = [x for x in vidNames if x.endswith(".mp4")]
# videoPath = '../training/download_train-val/validationFiles/'
# vidNames = os.listdir(videoPath)
# vidNames = [x for x in vidNames if x.endswith(".mp4")]
# vidNames.extend(vidNamesTrain)
# Initialize detectors, load it for face detection
predictorPath = 'coreData/shape_predictor_68_face_landmarks.dat'
faceDetector = dlib.get_frontal_face_detector()
shapePredictor = dlib.shape_predictor(predictorPath)
saveFetPath = 'tmpData/visualFetB/'
saveVidPath = 'tmpData/vidData/'
if not os.path.exists(saveFetPath):
os.makedirs(saveFetPath)
if not os.path.exists(saveVidPath):
os.makedirs(saveVidPath)
vidNames = vidNames
for i in range(len(vidNames)):
fileName = vidNames[i]
if (os.path.isfile(saveFetPath+fileName.strip('.mp4')+'.npy')):
continue
frameList = GetFrames(videoPath+fileName, redFact = 0.5, skipLength = 5)
savePath = saveVidPath + fileName.strip('.mp4')
# np.save(savePath, frameList)
# Do not save, too large!
faceList = DetectFaceLandmarksInList(frameList, faceDetector, shapePredictor)
savePath = saveFetPath + fileName.strip('.mp4')
np.save(savePath, faceList)
print ('\r'), ((i*(1.0))/len(vidNames)), 'part completed. Currently at file:', fileName,
sys.stdout.flush()
print '\n'
def __init__(self, visulize = True):
self.detector = dlib.get_frontal_face_detector()
self.predictor = dlib.shape_predictor(predictor_path)
self.fronter = getDefaultFrontalizer()
if visulize:
self.win = dlib.image_window()
self.win2 = dlib.image_window()
else:
self.win = self.win2 = None
def __init__(self, predictor_path):
"""
:rtype : LandmarkExtractor
"""
self.predictor_path = predictor_path
self.detector = dlib.get_frontal_face_detector()
self.predictor = dlib.shape_predictor(self.predictor_path)
self.featureMatrix = FeatureMatrix()
def __init__(self, facePredictor):
"""
Instantiate an 'AlignDlib' object.
:param facePredictor: The path to dlib's
:type facePredictor: str
"""
self.detector = dlib.get_frontal_face_detector()
self.predictor = dlib.shape_predictor(facePredictor)
def detect(self, src): if np.ndim(src) == 3: src = cv2.cvtColor(src, cv2.COLOR_BGR2GRAY) src = cv2.equalizeHist(src) detector = dlib.get_frontal_face_detector() rects = detector(src, 1) if len(rects) == 0: return [] return rects
def __init__(self):
"""Summary."""
model_filename = 'shape_predictor_68_face_landmarks.dat'
if not os.path.exists(model_filename):
os.system(
'wget http://sourceforge.net/projects/dclib/files/dlib/v18.10/shape_predictor_68_face_landmarks.dat.bz2')
os.system('bunzip2 shape_predictor_68_face_landmarks.dat.bz2')
self.model = dlib.shape_predictor(model_filename)
self.detector = dlib.get_frontal_face_detector()
def live_cam_detect(ss_path, sep_path, cc_path, usr_list):
"""
程序核心函数，主要利用opencv使用摄像头，然后调用上述函数进行完整的程序过程
:param usr_list: 从参数入口传入初始化获取的员工列表
:param ss_path: screenshots路径
:param sep_path: separate路径
:param cc_path: 缓存路径
"""
path_screenshots = ss_path
# 人脸检测模块
detector = dlib.get_frontal_face_detector()
# dlib同样支持轮廓预测
# predictor = dlib.shape_predictor(pre_path)
# 开启笔记本默认摄像头
cap = cv2.VideoCapture(0)
# 设置分辨率及格式
cap.set(3, 960)
if cap.isOpened(): # 检测摄像头是否开启
print("Camera is on.")
ss_cnt = 0 # 仅用于screenshot计数，图片起名字用。测试用
while cap.isOpened(): # 程序主循环，保持摄像头开启状态
flag, img_rd = cap.read() # 两个参数，
# 一个flag是接收读取图片状态，即True or False第二个是frame表示截取到的一帧的图片
k = cv2.waitKey(1) # 等待信号指令
img_gray = cv2.cvtColor(img_rd, cv2.COLOR_RGB2GRAY)
# 灰度测试并保存，用于人脸的探测
# 灰度有利于提高检测速度
faces = detector(img_gray, 0) # 利用灰度检查人脸
font = cv2.FONT_HERSHEY_SIMPLEX # 设置指令显示的字体
if k & 0xFF == ord('q'): # 输入q时退出循环
break
else: # 否则，开始干正事
# 开始画方框用于标识人脸
if len(faces) != 0:
for face in faces: # faces并非数字类型，其包含了众多实用的函数通过这些函数计算出矩形框的长宽
cv2.rectangle(img_rd, tuple([face.left(),
face.top()]),
tuple([face.right(),
face.bottom()]), (0, 255, 255), 2)
# 写入文字，显示总共显示的人脸数
cv2.putText(img_rd, "Faces in all: " + str(len(faces)),
(20, 350), font, 0.8, (255, 215, 0), 2,
cv2.LINE_AA)
else: # 如果人脸数为零则显示没有脸
cv2.putText(img_rd, "no face", (20, 350), font, 0.8,
(255, 215, 0), 2, cv2.LINE_AA)
# 注明指令和要求并设置字体格式 字体：HERSHEY_SIMPLEX 颜色：GOLD 粗细：2
img_rd = cv2.putText(img_rd, "Press 'S': Screen shot", (20, 400),
font, 0.8, (255, 215, 0), 2, cv2.LINE_AA)
img_rd = cv2.putText(img_rd, "Press 'C': Confirm image", (20, 450),
font, 0.8, (255, 215, 0), 2, cv2.LINE_AA)
img_rd = cv2.putText(img_rd, "Press 'Q': Quit", (20, 500), font,
0.8, (255, 215, 0), 2, cv2.LINE_AA)
img_rd = cv2.putText(img_rd, "Press 'O': Output", (20, 550), font,
0.8, (255, 215, 0), 2, cv2.LINE_AA)
img_rd = cv2.putText(
img_rd, time.strftime("%Y-%m-%d %H:%M:%S", time.localtime()),
(900, 50), font, 0.8, (255, 255, 0), 2, cv2.LINE_AA)
# s键保存截图，仅作为初期测试用功能
if k & 0xFF == ord('s'):
ss_cnt += 1
print(path_screenshots + "screenshot" + "_" + str(ss_cnt) + "_" +
time.strftime("%Y-%m-%d-%H-%M-%S", time.localtime()) +
".jpg")
cv2.imwrite(
path_screenshots + "screenshot" + "_" + str(ss_cnt) + "_" +
time.strftime("%Y-%m-%d-%H-%M-%S", time.localtime()) + ".jpg",
img_rd)
# c->confirm 确认录入信息，程序开始进行比对
if k & 0xFF == ord('c'):
print('\n', '*' * 50, '\n')
print("Face info confirmed, detecting ...")
cv2.imwrite(cc_path + "confirm_cache.jpg", img_rd) # 将截图写入缓存路径
faces = face_separate(cc_path, sep_path,
detector) # 从缓存路径读取原始图片并分割
print(faces, " faces detected. Recognizing...")
print("Please Wait ...")
stuff_list = [] # 员工列表
guest_list = [] # 客人列表
reco_cnt = 0 # 识别出的人脸数
unreco_cnt = 0 # 未识别出的人脸数
for face_num in range(0, faces):
read_path = sep_path + "img_face_" + str(
face_num + 1) + ".jpg" # 自动依次读取分割路径下的图片
result = face_recognize(online_client, read_path) # 开始上传检测
if result['error_msg'] == 'SUCCESS': # 如果检测成功
if result['result']['user_list'][0][
'score'] >= 80: # 相似度阀值定为80
stuff_list.append(
result['result']['user_list'][0]['user_id'])
reco_cnt += 1
rb = xlrd.open_workbook('./log/employee.xls')
wb = copy(rb)
ws = wb.get_sheet(0)
sheet1 = rb.sheet_by_index(0)
if sheet1.cell_value(
usr_list.index(stuff_list[reco_cnt - 1]) + 1,
1) == 'Registered':
wx.MessageBox("* * " + stuff_list[reco_cnt - 1] +
' has already checked in. * *',
caption='Ops!')
# print("* * ", stuff_list[reco_cnt - 1], ' has already checked in. * *')
else:
wx.MessageBox('* *Welcome ' +
stuff_list[reco_cnt - 1] +
'! Check in successful.* *',
caption='Congratulate!')
# print('* *Welcome ', stuff_list[reco_cnt - 1], '! Check in successful.* *')
# 打开已经初始化的表格准备修改状态
ws.write(
usr_list.index(stuff_list[reco_cnt - 1]) + 1, 1,
'Registered')
ws.write(
usr_list.index(stuff_list[reco_cnt - 1]) + 1, 2,
time.strftime("%Y-%m-%d %H:%M:%S",
time.localtime()))
wb.save('./log/employee.xls')
else: # 未达到80相似度，稍后询问是否手动录入
unreco_cnt += 1
guest_list.append(read_path)
print(reco_cnt, ' of ', faces,
'faces recognized. Process complete.')
# 如果存在未识别的人脸
if unreco_cnt != 0:
type_flag = 1
while type_flag: # type flag仅作为防止输入错误指令用
regist_flag = input(
"Unemployee faces detected. Register? (Y/N)\n")
if regist_flag == 'Y' or regist_flag == 'y':
for i in range(0, unreco_cnt):
show_img(guest_list[i]) # 显示未识别成员工的图片，请求用户指认该人脸姓名
name = input(
"Who is this(name)? Please watch the image on your right side.\n"
)
face_register(online_client, guest_list[i],
'employee', name) # 录入姓名
rb = xlrd.open_workbook('./log/employee.xls')
wb = copy(rb)
ws = wb.get_sheet(0)
# 在上传的同时更新本地出勤表
ws.write(len(usr_list) + 2, 0, name)
ws.write(len(usr_list) + 2, 1, 'Registered')
ws.write(
len(usr_list) + 2, 2,
time.strftime("%Y-%m-%d-%H-%M-%S",
time.localtime()))
wb.save('./log/employee.xls')
type_flag = 0
elif regist_flag == 'N' or regist_flag == 'n':
for j in range(0, unreco_cnt):
face_register(
online_client, guest_list[j], 'guest',
time.strftime("%Y%m%d%H%M%S",
time.localtime()))
type_flag = 0 # 如果录入成功则改变flag终止循环
else:
print('Invalid Command.')
print('\n', '*' * 50, '\n')
if k & 0xFF == ord('o'):
# 请求输入导出出勤表的路径
output_path = input(
"Please input output path (Backspace for default path):")
# 设置空格为默认路径
if output_path == ' ':
shutil.copy('./log/employee.xls', './')
print('File has been saved to ' +
os.path.dirname(os.path.abspath('./employee.xls')))
elif os.path.exists(output_path):
shutil.copy('./log/employee.xls', output_path)
print('File has been saved to ' + output_path)
else:
# 无效路径
print('Invalid path.')
# 设置摄像头窗口的相关参数
cv2.namedWindow("camera", 1)
cv2.imshow("camera", img_rd)
cap.release() # 释放摄像头
if not cap.isOpened(): # 检测摄像头状态
print("Camera is off.")
cv2.destroyAllWindows() # 关闭所有窗口
def calculate_distance(reference_clip, compare_clip):
# construct the `argument parse and parse the arguments
# ap = argparse.ArgumentParser()
# ap.add_argument("-p", "--shape-predictor", required=True, default = "shape_predictor_68_face_landmarks.dat",
# help="path to facial landmark predictor")
# ap.add_argument("-r", "--picamera", type=int, default=-1,
# help="whether or not the Raspberry Pi camera should be used")
# args = vars(ap.parse_args())
# initialize dlib's face detector (HOG-based) and then create
# the facial landmark predictor
print("[INFO] loading facial landmark predictor...")
# 얼굴 자체를 인식하는 기능
detector = dlib.get_frontal_face_detector()
# face 안에 얼굴 인식하는 기능
predictor = dlib.shape_predictor("shape_predictor_68_face_landmarks.dat")
# predictor = dlib.shape_predictor(args["shape_predictor"])
# vs = VideoStream(usePiCamera=args["picamera"] > 0).start()
# 비디오 출력 : https://076923.github.io/posts/Python-opencv-4/
# https://docs.opencv.org/master/dd/d43/tutorial_py_video_display.html
# capture = cv2.VideoCapture("cut_2.mp4")
clips = [reference_clip, compare_clip]
time.sleep(2.0)
clips_frame_info = []
for clip in clips:
i = 0
every_frame_info = []
# loop over the frames from the video stream
while True:
# grab the frame from the threaded video stream, resize it to
# have a maximum width of 400 pixels, and convert it to
# ret, frame = capture.read() # frame = numpy array임
frame = clip.get_frame(i * 1.0 / clip.fps)
i += skip_frame_rate # 1초에 60 fps가 있으므로 몇개는 skip해도 될거 같음!
if (i * 1.0 / clip.fps) > clip.duration:
break
# if not ret:
# print("Error")
# break
# width 높이면 더 판별 잘되지만, computational power 높음
# The benefit of increasing the resolution of the input image prior to face detection is that it may allow us to detect more faces in the imag
frame = imutils.resize(frame, width=800)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# detect faces in the grayscale frame
# 얼굴 자체의 위치를 찾음
rects = detector(gray, 0)
if len(rects) > 0:
# 얼굴 개수만큼 loop
max_width = 0
max_rect = None
# 얼굴 박스 제일 큰거 하나로
for rect in rects:
if int(rects[0].width()) > max_width:
max_rect = rect
# determine the facial landmarks for the face region, then
# convert the facial landmark (x, y)-coordinates to a NumPy array
shape = predictor(gray, max_rect)
shape = face_utils.shape_to_np(shape)
every_frame_info.append(shape)
else:
every_frame_info.append([])
clips_frame_info.append(np.array(every_frame_info))
cv2.destroyAllWindows()
min_size = min(len(clips_frame_info[0]), len(clips_frame_info[1]))
min_diff = float("inf")
min_idx = 0
for i in range(min_size):
if len(clips_frame_info[0][i]) > 0 and len(
clips_frame_info[1][i]) > 0: # 얼굴 둘다 있으면
# 양쪽 눈
left_eye = ((clips_frame_info[0][i][36][0] -
clips_frame_info[1][i][36][0])**2 +
(clips_frame_info[0][i][36][1] -
clips_frame_info[1][i][36][1])**2)**0.5
right_eye = ((clips_frame_info[0][i][45][0] -
clips_frame_info[1][i][45][0])**2 +
(clips_frame_info[0][i][45][1] -
clips_frame_info[1][i][45][1])**2)**0.5
total_diff = left_eye + right_eye
if min_diff > total_diff:
min_diff = total_diff
min_idx = i
return min_diff, (min_idx *
skip_frame_rate) / clip.fps # 거리와 해당 초 위치를 계산해준다!
def getSkeleton(frame, isHarr=True):
### Inferencing Pose ###
nPoints = 18
POSE_PAIRS = [ [1,0],[1,2],[1,5],[2,3],[3,4],[5,6],\
[6,7],[1,8],[8,9],[9,10], [14,11], [11,12], [12,13] ]
BODY_PARTS = {
"Nose": 0,
"Neck": 1,
"RShoulder": 2,
"RElbow": 3,
"RWrist": 4,
"LShoulder": 5,
"LElbow": 6,
"LWrist": 7,
"RHip": 8,
"RKnee": 9,
"RAnkle": 10,
"LHip": 11,
"LKnee": 12,
"LAnkle": 13,
"REye": 14,
"LEye": 15,
"REar": 16,
"LEar": 17,
"Background": 18
}
idx_2_BODY_PARTS = {int(num): part for part, num in BODY_PARTS.items()}
frameCopy = np.copy(frame)
frameWidth = frame.shape[1]
frameHeight = frame.shape[0]
threshold = 0.1
# inHeight / inWidth
inWidth = 368
inHeight = 368
inpBlob = cv2.dnn.blobFromImage(frame, 1.0/255, (inWidth, inHeight),\
(0,0,0), swapRB=False, crop=False)
dotsize = int((frameWidth + frameHeight) / (inWidth))
textsize = int((frameWidth + frameHeight) / (inWidth * 4))
linesize = dotsize // 2
network.setInput(inpBlob)
output = network.forward()
H = output.shape[2]
W = output.shape[3]
points = []
points_with_num = []
facelist = []
### Inferencing face & emotion ###
if isHarr:
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
facelist = face_cascade.detectMultiScale(gray, 1.3, 5)
facelist = list(np.array(facelist).flatten())[:4]
else:
import dlib
hogFaceDetector = dlib.get_frontal_face_detector()
faceRects = hogFaceDetector(frame, 0)
facelist = [faceRects[0].left(),faceRects[0].top(),\
faceRects[0].right()-faceRects[0].left(), faceRects[0].bottom()-faceRects[0].top()]
emotion = 'None' # 초기값 None 이 아니라 string 이어야 합니다.
if len(facelist) != 0:
(fx, fy, fw, fh) = list(map(int, facelist))
# print(type(fx),type(fy),type(fw),type(fh))
face = cv2.cvtColor(frame[fy:fy + fh, fx:fx + fw], cv2.COLOR_BGR2GRAY)
face = cv2.resize(face, dsize=(48, 48), interpolation=cv2.INTER_LINEAR)
input_face = face.reshape(-1, 48, 48, 1)
model = load_model()
label_dict = {
'Angry': 0,
'Disgust': 1,
'Fear': 2,
'Happy': 3,
'Neutral': 4,
'Sad': 5,
'Surprise': 6
}
label_dict = {num: emotion for emotion, num in label_dict.items()}
emotion = label_dict[np.argmax(model.predict(input_face))]
## memory flush ##
### Draw points, lines, rectangle ###
### Pose estimation ###
for i in range(nPoints):
# confidence map of corresponding body's part
probMap = output[0, i, :, :]
minVal, prob, minLoc, point = cv2.minMaxLoc(probMap)
x = (frameWidth * point[0]) / W
y = (frameHeight * point[1]) / H
if prob > threshold:
cv2.circle(frameCopy, (int(x), int(y)), dotsize, (0,255,255),\
thickness=dotsize, lineType=cv2.FILLED)
cv2.putText(frameCopy, "{}".format(i), (int(x), int(y)),\
cv2.FONT_HERSHEY_SIMPLEX, textsize, (0,0,255), textsize, \
lineType=cv2.LINE_AA)
cv2.circle(frame, (int(x), int(y)), dotsize, (0, 0, 255),\
thickness=dotsize, lineType=cv2.FILLED)
# Add the point to the list if the probability is greater than the threshold
points.append((int(x), int(y)))
points_with_num.append(
(int(x), int(y), int(i), idx_2_BODY_PARTS[int(i)]))
else:
points.append(None)
# Draw Skeleton
for pair in POSE_PAIRS:
partA = pair[0]
partB = pair[1]
if points[partA] and points[partB]:
cv2.line(frame, points[partA], points[partB], (0, 255, 255),
linesize)
### Draw Face box and emotion ###
if len(facelist) != 0:
cv2.rectangle(frame, (int(fx), int(fy)), (int(fx + fw), int(fy + fh)),
(0, 255, 0), linesize)
cv2.rectangle(frameCopy, (int(fx), int(fy)),
(int(fx + fw), int(fy + fh)), (0, 255, 0), linesize)
img_skeleton = frame
img_with_dot = frameCopy
return img_skeleton, img_with_dot, points_with_num, emotion
def __init__(self, shape_predictor_path):
self.detector = dlib.get_frontal_face_detector()
self.shape_predictor = dlib.shape_predictor(shape_predictor_path)
def main():
global COUNTER
global TOTAL
shape_predictor_path = '../shape_predictor_68_face_landmarks.dat'
video_path = '../blink_test.mp4'
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(shape_predictor_path)
print('Load detector successfully')
(lStart, lEnd) = face_utils.FACIAL_LANDMARKS_IDXS["left_eye"]
(rStart, rEnd) = face_utils.FACIAL_LANDMARKS_IDXS["right_eye"]
vs = FileVideoStream(video_path).start()
file_stream = True
while True:
if file_stream and not vs.more():
break
frame = vs.read()
frame = imutils.resize(frame, width=450)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# detect faces in the grayscale frame
rects = detector(gray, 0)
for rect in rects:
# determine the facial landmarks for the face region, then
# convert the facial landmark (x, y)-coordinates to a NumPy
# array
shape = predictor(gray, rect)
shape = face_utils.shape_to_np(shape)
# extract the left and right eye coordinates, then use the
# coordinates to compute the eye aspect ratio for both eyes
leftEye = shape[lStart:lEnd]
rightEye = shape[rStart:rEnd]
leftEAR = eye_aspect_ratio(leftEye)
rightEAR = eye_aspect_ratio(rightEye)
# average the eye aspect ratio together for both eyes
ear = (leftEAR + rightEAR) / 2.0
leftEyeHull = cv2.convexHull(leftEye)
rightEyeHull = cv2.convexHull(rightEye)
cv2.drawContours(frame, [leftEyeHull], -1, (0, 255, 0), 1)
cv2.drawContours(frame, [rightEyeHull], -1, (0, 255, 0), 1)
if ear < EYE_AR_THRESH:
COUNTER += 1
else:
if COUNTER >= EYE_AR_CONSEC_FRAMES:
TOTAL += 1
COUNTER = 0
cv2.putText(frame, "Blinks: {}".format(TOTAL), (10, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
cv2.putText(frame, "EAR: {:.2f}".format(ear), (300, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
cv2.imshow("Frame", frame)
key = cv2.waitKey(1) & 0xFF
if key == ord("q"):
break
cv2.destroyAllWindows()
vs.stop()
class OrientationClassifier(object):
# Initializing the face detector
detector = dlib.get_frontal_face_detector()
def __init__(self, args):
"""
construct the argument parser and parse the arguments
args: parsed arguments
"""
try:
# reading the first image
self.image = cv2.imread(args["image"])
# creating another copy of first image for comparison. After the facial detector has been applied, if the image exist in correct orientation,
# it will have a bounding box and hence would be different. Else, it would be the same image
self.image2 = cv2.imread(args["image"])
# using predictor for identification of facial landmarks in the face
self.predictor = dlib.shape_predictor(args["shape_predictor"])
self.imageResize()
except:
raise Exception
def imageResize(self):
"""
Construct for resizing the image
"""
self.image = imutils.resize(self.image, width=500)
self.image2 = imutils.resize(self.image2, width=500)
def getDifference(self):
"""
Construct for detection of facial landmark and thus identifying if the imaghe is in correct orientation not
"""
try:
# using the dlib detector. This is a general detector used for detection of regid objects.
rects = self.detector(self.image, 1)
for (i, rect) in enumerate(rects):
# determine the facial landmarks for the face region, then
# convert the facial landmark (x, y)-coordinates to a NumPy
# array
shape = self.predictor(self.image, rect)
shape = face_utils.shape_to_np(shape)
# convert dlib's rectangle to a OpenCV-style bounding box
# [i.e., (x, y, w, h)], then draw the face bounding box
(x, y, w, h) = face_utils.rect_to_bb(rect)
cv2.rectangle(self.image, (x, y), (x + w, y + h), (0, 255, 0),
2)
# show the face number
cv2.putText(self.image, "Face #{}".format(i + 1),
(x - 10, y - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5,
(0, 255, 0), 2)
# loop over the (x, y)-coordinates for the facial landmarks
# and draw them on the image
for (x, y) in shape:
cv2.circle(self.image, (x, y), 1, (0, 0, 255), -1)
# Getting difference between altered images and initial image to see if there is a bounding box present.
#If bounding box is present, the image is in correct oreintation
difference = cv2.subtract(self.image2, self.image)
b, g, r = cv2.split(difference)
if cv2.countNonZero(b) == 0 and cv2.countNonZero(
g) == 0 and cv2.countNonZero(r) == 0:
return 1
else:
return 0
except:
# This is because even a non image file can be read, however, it would raise an error when applying a detector
print('Please check the image entered and provide a valid image')
return -1
def fixOrientation(self):
"""
construct for fixing orientation and identifying the class
"""
count = 1
classV = 0
# we loop through 0 to 360 until the facial landmarks are detecting in the image
for angle in np.arange(0, 360, 90):
count += 1
# rotating both images because we want to have both images similar and the only difference to be the presence
# or absense of facial landmarks
self.image = imutils.rotate_bound(self.image, angle)
self.image2 = imutils.rotate_bound(self.image2, angle)
self.imageResize()
val = self.getDifference()
if val == 0:
if count == 2:
classV = 4
elif count == 3:
classV = 3
elif count == 4:
classV = 2
return self.image2, classV
def findFace(l, s, f, v, a):
#Create HOG face detector from dlib class
face_detector = dlib.get_frontal_face_detector()
landmark_predictor_model = "shape_predictor_68_face_landmarks.dat"
face_pose_predictor = dlib.shape_predictor(landmark_predictor_model)
video_cap = cv2.VideoCapture(0) #image from video (default webcam)
#if no one is in the frame
last = True
(leftStart, leftEnd) = facial_locations["left_eye"]
(rightStart, rightEnd) = facial_locations["right_eye"]
EYE_THRESH = 0.3
EYE_FRAMES = 3
COUNTER = 0
TOTAL = 0
while True:
#capture frame by frame
#returns return code(tells us if we have run out of frames)
#frame = one frame
ret, frame = video_cap.read()
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
detected_faces = face_detector(gray, 1)
if (v):
print("Found {} faces in this frame.".format(len(detected_faces)))
if (len(detected_faces) == 0): last = True
# Loop through all faces found
for i, face_rect in enumerate(detected_faces):
if (v):
print(
"- Face #{} found at Left: {} Top: {} Right: {} Bottom: {}"
.format(i, face_rect.left(), face_rect.top(),
face_rect.right(), face_rect.bottom()))
# Draw a box around each face we found
cv2.rectangle(frame, (face_rect.left(), face_rect.top()),
(face_rect.right(), face_rect.bottom()), (0, 255, 0),
2)
crop = frame[face_rect.top():face_rect.bottom(),
face_rect.left():face_rect.right()]
#Generate and store landmarks
face_landmarks = face_pose_predictor(frame, face_rect)
points = store_landmarks(face_landmarks)
if (a):
#Collect facial feature coordinates
leftEye = points[leftStart:leftEnd]
rightEye = points[rightStart:rightEnd]
lear = eye_aspect_ratio(leftEye)
rear = eye_aspect_ratio(rightEye)
aear = (lear + rear) / 2.0
leftEyeHull = cv2.convexHull(leftEye)
rightEyeHull = cv2.convexHull(rightEye)
cv2.drawContours(frame, [leftEyeHull], -1, (0, 255, 0), 1)
cv2.drawContours(frame, [rightEyeHull], -1, (0, 255, 0), 1)
if aear < EYE_THRESH: COUNTER += 1
else:
if COUNTER >= EYE_FRAMES: TOTAL += 1
COUNTER = 0
cv2.putText(frame, "Blinks: {}".format(TOTAL), (10, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
#If user wants to save faces, save into file
#TODO: given filename save
if (s and last):
last = False
cv2.imwrite("face.jpg", crop)
#If user wants to show landmarks on screen
if (l):
frame = show_landmarks(frame, face_landmarks, points, v)
#Display info
cv2.putText(frame, 'SEARCHING FOR FACES...', topLeft, font, fontScale,
fontColor, lineType)
cv2.imshow('Video', frame)
# Close when 'q' key is pressed
if cv2.waitKey(1) & 0xFF == ord('q'):
break
video_cap.release()
cv2.destroyAllWindows()
if w1 > w2:
width = w2
else:
width = w1
frameV1 = cv2.resize(frameV1, (width, height))
frameV2 = cv2.resize(frameV2, (width, height))
fps1 = video1.get(cv2.CAP_PROP_FPS)
fps2 = video1.get(cv2.CAP_PROP_FPS)
fourcc = cv2.VideoWriter_fourcc(*'XVID')
vOut1 = cv2.VideoWriter('output1.mp4', cv2.VideoWriter_fourcc(*'MP4V'), fps1,
(width, height))
vOut2 = cv2.VideoWriter('output2.mp4', cv2.VideoWriter_fourcc(*'MP4V'), fps2,
(width, height))
#feature detection and extraction
DETECTOR = dlib.get_frontal_face_detector()
PREDICTOR = dlib.shape_predictor(PREDICTOR_PATH)
detectedPts1 = DETECTOR(frameV1, 1)
detectedPts2 = DETECTOR(frameV2, 2)
pointsF1 = []
pointsF2 = []
frameNum = 0
for k, d in enumerate(detectedPts2):
shape = PREDICTOR(frameV2, d)
for i in range(68):
pointsF2.append((shape.part(i).x, shape.part(i).y))
for k, d in enumerate(detectedPts1):
shape = PREDICTOR(frameV1, d)
for i in range(68):
pointsF1.append((shape.part(i).x, shape.part(i).y))
old2FeaturePoints = np.array(pointsF2)
# if len(sys.argv) != 2:
# print(
# "Call this program like this:\n"
# " ./face_jitter.py shape_predictor_5_face_landmarks.dat\n"
# "You can download a trained facial shape predictor from:\n"
# " http://dlib.net/files/shape_predictor_5_face_landmarks.dat.bz2\n")
# exit()
predictor_path = "E:/bigdata/ai/dlib/data/shape_predictor_5_face_landmarks.dat" #sys.argv[1]
face_file_path = "./faces/Tom_Cruise_avp_2014_4.jpg"
# Load all the models we need: a detector to find the faces, a shape predictor
# to find face landmarks so we can precisely localize the face
detector = dlib.get_frontal_face_detector()
sp = dlib.shape_predictor(predictor_path)
# Load the image using dlib
img = dlib.load_rgb_image(face_file_path)
# Ask the detector to find the bounding boxes of each face.
dets = detector(img)
num_faces = len(dets)
# Find the 5 face landmarks we need to do the alignment.
faces = dlib.full_object_detections()
for detection in dets:
faces.append(sp(img, detection))
class FrameHandler(object):
"""Handle all necessary transformations of a frame to predict
the facial expressions.
To simplify the process and to achieve consistent results, this class
only focusses on one single face. If there are multiple faces in the frame,
only the first detected face will be returned.
Detecting multiple faces would require we track where in subsequent frames
a persons face is, and make sure the data is aligned with the faces.
Our concept for this assignment doesn't involve multiple faces, so this
functionality is forsaken in favour of a single face.
Class attributes:
predictor: The dlib facial landmarks shape predictor
detector: The dlib face detector
clahe: The cv2 CLAHE algorithm
Attributes:
detection: The detected face from the original frame
resized_frame: A new frame, resized from the original frame
based on the detected face.
resized_detection: The detections of the face from the
resized frame.
"""
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(
'models/shape_predictor_68_face_landmarks.dat')
def __init__(self, frame):
self.frame = frame
gray = cv2.cvtColor(self.frame, cv2.COLOR_BGR2GRAY)
self.clahe_image = self.clahe.apply(gray)
def _resize_face(self, rect):
"""Resize the given face to 350x350 (same as dataset)
Args:
rect: The rectangle of the face. This is the area of the frame
that will be cut from the frame.
Returns:
frame | None: A resized version of the `self.clahe_image`,
or None is resizing failed.
"""
x = rect.left()
y = rect.top()
w = rect.right() - x
h = rect.bottom() - y
cut = self.clahe_image[y:y + h, x:x + w]
try:
return cv2.resize(cut, (350, 350))
except:
return None
def get_vectorized_landmarks(self, resized=True):
"""Get the vectorized landmarks of the frame.
Args:
resized (bool): If true, gets the landmarks from the resized
frames, otherwise, use the original frame.
"""
if resized:
frame = self.resized_frame
rect = self.resized_detection
else:
frame = self.clahe_image
rect = self.detection
if rect is None:
return None
shape = self.predictor(frame, rect)
xlist = []
ylist = []
for i in range(68):
xlist.append(float(shape.part(i).x))
ylist.append(float(shape.part(i).y))
xmean = np.mean(xlist)
ymean = np.mean(ylist)
xcentral = [(x - xmean) for x in xlist]
ycentral = [(y - ymean) for y in ylist]
landmarks_vectorised = []
for x, y, w, z in zip(xcentral, ycentral, xlist, ylist):
landmarks_vectorised.append(w)
landmarks_vectorised.append(z)
meannp = np.asarray((ymean, xmean))
coornp = np.asarray((z, w))
dist = np.linalg.norm(coornp - meannp)
landmarks_vectorised.append(dist)
landmarks_vectorised.append(
(math.atan2(y, x) * 360) / (2 * math.pi))
return landmarks_vectorised
def draw_landmarks(self, thickness=1):
# Can't draw that which doesn't exist.
if self.detections is None:
return
shape = self.predictor(self.clahe_image, self.detection)
for i in range(68):
cv2.circle(self.frame, (shape.part(i).x, shape.part(i).y),
1, (0, 0, 255),
thickness=thickness)
@property
def detection(self):
if not hasattr(self, '_detections'):
self._detections = self.detector(self.clahe_image, 1)
return self._detections[0] if len(self._detections) > 0 else None
@property
def resized_frame(self):
if not hasattr(self, '_resized'):
if self.detection:
self._resized = self._resize_face(self.detection)
else:
self._resized = None
return self._resized
@property
def resized_detection(self):
if not hasattr(self, '_re_detections'):
if self.resized_frame is not None:
self._re_detections = self.detector(self.resized_frame, 1)
else:
self._re_detections = []
return self._re_detections[0] if len(self._re_detections) > 0 else None
def FrameExtract(path1, path2):
print(f"Working with video file {path1[path1.find('/')+1:]} ")
video_obj = cv2.VideoCapture(path1) #Video object
detector = dlib.get_frontal_face_detector() #Face detector
predictor = dlib.shape_predictor(
path2
) #Landmark identifier. Set the filename to whatever you named the downloaded file
count = 0
ret = 1
p = []
while ret: #Runs ONCE for each FRAME
ret, frame = video_obj.read()
if (ret == 0):
continue
gray = cv2.cvtColor(frame,
cv2.COLOR_BGR2GRAY) #Convert image to grayscale
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
clahe_image = clahe.apply(gray)
detections = detector(clahe_image, 1)
#Detect the faces in the image
for k, d in enumerate(detections):
#print("enter loop") #For each detected face
shape = predictor(clahe_image, d)
#shape1 = face_utils.shape_to_np(shape) This works by utilizing imutis library
#print(shape1)
#print(shape.part(i).y)#Get coordinates
#print(shape)
vec = np.empty([68, 2], dtype=int)
for b in range(68):
vec[b][0] = shape.part(b).x
vec[b][1] = shape.part(b).y
#print(vec.shape)
v = vec.tolist()
#print(type(v))
#print(type(p))
p.append(v)
#
#print(shape1-vec)
#cv2.imshow("image", frame) #Display the frame
#print("enter after loop")
if not os.path.exists(path1[:path1.find(".")]):
os.makedirs(path1[:path1.find(".")])
#print(count)
#cv2.imwrite(path1[:path1.find(".")]+"/frame%d.jpg" % count, imag)
count += 1
p1 = np.array(p)
print(p1.shape)
panel = pd.Panel(
p1,
items=['Frame {}'.format(i) for i in range(0, p1.shape[0])],
major_axis=['Landmark {}'.format(i) for i in range(0, 68)],
minor_axis=['x', 'y'])
pnl = panel.to_frame()
print(p1)
print(pnl)
pnl.to_csv(path1[:path1.find(".")] + '/landmark_points_dataframe.csv')
np.save(path1[:path1.find(".")] + '/landmark_points_array.out', p1)
#np.savetxt(path1[:path1.find(".")]+"/reshaped.txt", p1.reshape((3,-1)), fmt="%s", header=str(p1.shape))
return p1
def run_detector():
# define two constants, one for the eye aspect ratio to indicate
# blink and then a second constant for the number of consecutive
# frames the eye must be below the threshold for to set off the
# alarm
EYE_AR_THRESH = 0.3
EYE_AR_CONSEC_FRAMES = 48
# initialize the frame counter as well as a boolean used to
# indicate if the alarm is going off
COUNTER = 0
ALARM_ON = False
# initialize dlib's face detector (HOG-based) and then create
# the facial landmark predictor
print("[INFO] loading facial landmark predictor...")
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor("shape_predictor_68_face_landmarks.dat")
# grab the indexes of the facial landmarks for the left and
# grab the indexes of the facial landmarks for the left and
# right eye, respectively
(lStart, lEnd) = face_utils.FACIAL_LANDMARKS_IDXS["left_eye"]
(rStart, rEnd) = face_utils.FACIAL_LANDMARKS_IDXS["right_eye"]
# start the video stream thread
print("[INFO] starting video stream thread...")
# vs = VideoStream(src=args["webcam"]).start()
# # time.sleep(1.0)
cam = cv2.VideoCapture(1 + cv2.CAP_DSHOW)
# Current time in millis
start_time = int(round(time.time() * 1000))
#time 30 seconds later in millis
stop_time = int(round(time.time() * 1000)) + 5000
# loop over frames from the video stream
while True:
# grab the frame from the threaded video file stream, resize
# it, and convert it to grayscale
# channels)
ret, frame = cam.read()
frame = imutils.resize(frame, width=450)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# detect faces in the grayscale frame
rects = detector(gray, 0)
# loop over the face detections
for rect in rects:
# determine the facial landmarks for the face region, then
# convert the facial landmark (x, y)-coordinates to a NumPy
# array
shape = predictor(gray, rect)
shape = face_utils.shape_to_np(shape)
# extract the left and right eye coordinates, then use the
# coordinates to compute the eye aspect ratio for both eyes
leftEye = shape[lStart:lEnd]
rightEye = shape[rStart:rEnd]
leftEAR = eye_aspect_ratio(leftEye)
rightEAR = eye_aspect_ratio(rightEye)
# average the eye aspect ratio together for both eyes
ear = (leftEAR + rightEAR) / 2.0
# compute the convex hull for the left and right eye, then
# visualize each of the eyes
leftEyeHull = cv2.convexHull(leftEye)
rightEyeHull = cv2.convexHull(rightEye)
cv2.drawContours(frame, [leftEyeHull], -1, (0, 255, 0), 1)
cv2.drawContours(frame, [rightEyeHull], -1, (0, 255, 0), 1)
# check to see if the eye aspect ratio is below the blink
# threshold, and if so, increment the blink frame counter
if ear < EYE_AR_THRESH:
COUNTER += 1
# if the eyes were closed for a sufficient number of
# then sound the alarm
if COUNTER >= EYE_AR_CONSEC_FRAMES:
# if the alarm is not on, turn it on
if not ALARM_ON:
ALARM_ON = True
# check to see if an alarm file was supplied,
# and if so, start a thread to have the alarm
# sound played in the background
t = Thread(target=sound_alarm, args=("alarm.wav", ))
t.deamon = True
t.start()
# draw an alarm on the frame
cv2.putText(frame, "DROWSINESS ALERT!", (10, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
# otherwise, the eye aspect ratio is not below the blink
# threshold, so reset the counter and alarm
else:
COUNTER = 0
ALARM_ON = False
# draw the computed eye aspect ratio on the frame to help
# with debugging and setting the correct eye aspect ratio
# thresholds and frame counters
cv2.putText(frame, "EAR: {:.2f}".format(ear), (300, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
# show the frame
cv2.imshow("Frame", frame)
cv2.waitKey(1) & 0xFF
# Time to run detector for
# Currently set to 1 seconds for testing
start_time = int(round(time.time() * 1000))
if start_time > stop_time:
break
# do a bit of cleanup
cv2.destroyAllWindows()
cam.release()
def detect_faces(image):
face_detector = dlib.get_frontal_face_detector()
detected_faces = face_detector(image, 1)
face_frames = [(x.left(), x.top(), x.right(), x.bottom())
for x in detected_faces]
return face_frames
def main() :
args = vars(ap.parse_args())
EYE_AR_THRESH = args['threshold']
EYE_AR_CONSEC_FRAMES = args['frames']
COUNTER = 0
TOTAL = 0
print("[INFO] loading facial landmark predictor...")
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(args["shape_predictor"])
(lStart, lEnd) = face_utils.FACIAL_LANDMARKS_IDXS["left_eye"]
(rStart, rEnd) = face_utils.FACIAL_LANDMARKS_IDXS["right_eye"]
print("[INFO] starting video stream thread...")
print("[INFO] print q to quit...")
if args['video'] == "camera":
vs = VideoStream(src=1).start()
fileStream = False
else:
vs = FileVideoStream(args["video"]).start()
fileStream = True
time.sleep(1.0)
while True:
if fileStream and not vs.more():
break
frame = vs.read()
frame = imutils.resize(frame, width=450)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
rects = detector(gray, 0)
for rect in rects:
shape = predictor(gray, rect)
shape = face_utils.shape_to_np(shape)
leftEye = shape[lStart:lEnd]
rightEye = shape[rStart:rEnd]
leftEAR = eye_aspect_ratio(leftEye)
rightEAR = eye_aspect_ratio(rightEye)
ear = (leftEAR + rightEAR) / 2.0
leftEyeHull = cv2.convexHull(leftEye)
rightEyeHull = cv2.convexHull(rightEye)
cv2.drawContours(frame, [leftEyeHull], -1, (0, 255, 0), 1)
cv2.drawContours(frame, [rightEyeHull], -1, (0, 255, 0), 1)
if ear < EYE_AR_THRESH:
COUNTER += 1
else:
if COUNTER >= EYE_AR_CONSEC_FRAMES:
TOTAL += 1
COUNTER = 0
cv2.putText(frame, "Blinks: {}".format(TOTAL), (10, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
cv2.putText(frame, "EAR: {:.2f}".format(ear), (300, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
cv2.imshow("Frame", frame)
key = cv2.waitKey(1) & 0xFF
if key == ord("q"):
break
cv2.destroyAllWindows()
vs.stop()
import cv2 # imports OpenCV library and associated functions and methods
import numpy as np # imports numpy as "np" library and associated functions and methods
import dlib # imports dlib library and associated functions
path = "enter the path to the .dat file" #creates a variable with the address of the dlib library
predictor = dlib.shape_predictor(path) # creates a variable that stores the method of face decection from dlib library
detector = dlib.get_frontal_face_detector() # creates a variable that stores another method that dlib provides for detecting faces
def get_landmarks(im): # creating a function that gets the landmarks on the face to make the code easier (detector returns cordinates of major facial features)
rects = detector(im,1) # creates a variable that returns the number of faces in the screen
if len(rects) > 1: # To check if there are more than 1 faces, if yes, return an empty matrix
print("Toomanyfaces")
return np.matrix([0])
if len(rects) == 0: # To check if there are 0 faces, if yes, return an empty matrix
print("Toofewfaces")
return np.matrix([0])
return np.matrix([[p.x, p.y] for p in predictor(im, rects[0]).parts()]) #if there is exactly 1 face, the predictor method is used to append cordinates of facial features in a numpy matrix
def upper_lip(landmarks): # To find out the mean cordinates of the upper lip position
top_lip = [] # Creates a new list
for i in range(50,53): # Creating a range from "50" to "53" because they represent the upper lip in dlib face predictor
top_lip.append(landmarks[i]) # Appending the same in the top lip list created above
for j in range(61,64): # Creating a range from "61" to "64" because they represent the upper lip in dlib face predictor
top_lip.append(landmarks[j])
def blinkCounter(video):
directions = []
TOTAL = 0
COUNTER = 0
Total_Frames = 0
time_for_blink = []
consecutive_frames = []
print("starting blink detection on: ", video)
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(data_set)
(lStart, lEnd) = face_utils.FACIAL_LANDMARKS_IDXS["left_eye"]
(rStart, rEnd) = face_utils.FACIAL_LANDMARKS_IDXS["right_eye"]
vs = FileVideoStream(video).start()
fileStream = True
time.sleep(1.0)
while in_loop:
if fileStream and not vs.more():
return TOTAL, Total_Frames, time_for_blink, directions
frame = vs.read()
Total_Frames += 1
frame = imutils.resize(frame, width=450)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
rects = detector(gray, 0)
for rect in rects:
shape = predictor(gray, rect)
shape = face_utils.shape_to_np(shape)
leftEye = shape[lStart:lEnd]
rightEye = shape[rStart:rEnd]
leftEAR = eye_aspect_ratio(leftEye)
rightEAR = eye_aspect_ratio(rightEye)
ear = (leftEAR + rightEAR) / 2.0
if ear < EYE_AR_THRESH:
COUNTER += 1
consecutive_frames.append(Total_Frames)
else:
if COUNTER >= EYE_AR_CONSEC_FRAMES:
TOTAL += 1
if len(consecutive_frames) != 0:
timeStamp = consecutive_frames[0] / 30
time_for_blink.append(int(timeStamp))
consecutive_frames = []
COUNTER = 0
cv2.putText(frame, "Blinks: {}".format(TOTAL), (10, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
cv2.putText(frame, "EAR: {:.2f}".format(ear), (300, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
cv2.imshow("Frame", frame)
key = cv2.waitKey(1) & 0xFF
if key == ord("q"):
return TOTAL, Total_Frames, time_for_blink, directions
if key == ord("7"):
print("Upper Left")
save_position("UL", Total_Frames, directions)
if key == ord("8"):
print("Up")
save_position("U", Total_Frames, directions)
if key == ord("9"):
print("Upper Right")
save_position("UR", Total_Frames, directions)
if key == ord("4"):
print("left")
save_position("L", Total_Frames, directions)
if key == ord("5"):
print("middle")
save_position("M", Total_Frames, directions)
if key == ord("6"):
print("right")
save_position("R", Total_Frames, directions)
if key == ord("1"):
print("down left")
save_position("DL", Total_Frames, directions)
if key == ord("2"):
print("down")
save_position("D", Total_Frames, directions)
if key == ord("3"):
print("down right")
save_position("DR", Total_Frames, directions)
cv2.destroyAllWindows()
vs.stop()
def __init__(self, dlib_model):
self.face_regressor = dlib.shape_predictor(dlib_model)
self.face_detector = dlib.get_frontal_face_detector()
def get_cropped_image(self):
face = self.find_face(self.image) #[1][0]
print(face[0])
if face[0] is False:
from imutils import face_utils
import dlib
face_detector = dlib.get_frontal_face_detector()
rects = face_detector(self.image)
left, right, top, bottom = rects[0][0], rects[0][2], rects[0][
1], rects[0][3]
else:
face = face[1][0]
left, top, right, bottom = face[0], face[
1], face[0] + face[2], face[1] + face[3]
image_size = self.image.shape
hori_adjustment_scale = image_size[1] / 10
vert_adjustment_scale = image_size[0] / 10
# print('hori_adjustment_scale- {}'.format(hori_adjustment_scale))
# print('vert_adjustment_scale- {}'.format(vert_adjustment_scale))
#
# print('top- {}'.format(top))
# print('bottom- {}'.format(bottom))
# print('left- {}'.format(left))
# print('right- {}'.format(right))
#
if (left - hori_adjustment_scale) >= 0:
left -= hori_adjustment_scale
left = int(left)
else:
left = 0
if (right + hori_adjustment_scale) <= image_size[1]:
right += hori_adjustment_scale
right = int(right)
else:
right = image_size[1]
if (top - vert_adjustment_scale) >= 0:
print('ye')
top -= vert_adjustment_scale
top = int(top)
else:
top = 0
if (bottom + vert_adjustment_scale) <= image_size[0]:
bottom += vert_adjustment_scale
bottom = int(bottom)
else:
bottom = image_size[0]
# print('top- {}'.format(top))
# print('bottom- {}'.format(bottom))
# print('left- {}'.format(left))
# print('right- {}'.format(right))
self.image = self.image[top:bottom, left:right]
self.image = cv2.resize(self.image, (640, 490))
return self.image
def fatigue_detection(self,Adafruit_CharLCD):
#眼睛长宽比的阈值，如果超过这个值就代表眼睛长/宽大于采集到的平均值，默认已经"闭眼"
eyesRatioLimit=0
#数据采集的次数
collectCountInterval=40
#数据采集的计数，采集collectCountInterval次然后取平均值
collectCount=0
#用于数据采集的求和
collectSum=0
#是否开始检测
startCheck=False
#统计"闭眼"的次数
eyesCloseCount=0
#初始化dlib
detector=dlib.get_frontal_face_detector()
predictor=dlib.shape_predictor("shape_predictor_68_face_landmarks.dat")
#获取面部各器官的索引
#左右眼
(left_Start,left_End)=face_utils.FACIAL_LANDMARKS_IDXS["left_eye"]
(right_Start,right_End)=face_utils.FACIAL_LANDMARKS_IDXS["right_eye"]
# #嘴
# (leftMouth,rightMouth)=face_utils.FACIAL_LANDMARKS_IDXS['mouth']
# #下巴
# (leftJaw,rightJaw)=face_utils.FACIAL_LANDMARKS_IDXS['jaw']
# #鼻子
# (leftNose,rightNose)=face_utils.FACIAL_LANDMARKS_IDXS['nose']
# #左右眉毛
# (left_leftEyebrow,left_rightEyebrow)=face_utils.FACIAL_LANDMARKS_IDXS['left_eyebrow']
# (right_leftEyebrow,right_rightEyebrow)=face_utils.FACIAL_LANDMARKS_IDXS['right_eyebrow']
#开启视频线程，延迟
print("[INFO] starting video stream...")
cap = cv2.VideoCapture(0)
#调节摄像头帧率
cap.set(cv2.CAP_PROP_FPS,2)
#调整摄像头分辨率
# ret = cap.set(cv2.CAP_PROP_FRAME_WIDTH,240)
# ret = cap.set(cv2.CAP_PROP_FRAME_HEIGHT,130)
#初始化
print("[INFO] initializing...")
time.sleep(2.0)
#循环检测
while True:
#对每一帧进行处理，设置宽度并转化为灰度图
ret,frame = cap.read()
#print('collect data')
frame = imutils.resize(frame, width=320)
#转换为GRAY颜色空间
img = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
#检测灰度图中的脸
faces = detector(img, 0)
for k in faces:
#确定面部区域的面部特征点，将特征点坐标转换为numpy数组
shape = predictor(img, k)
shape = face_utils.shape_to_np(shape)
#左右眼
leftEye = shape[left_Start:left_End]
rightEye = shape[right_Start:right_End]
leftEyesVal = self.eyesRatio(leftEye)
rightEyesVal = self.eyesRatio(rightEye)
#凸壳
leftEyeHull = cv2.convexHull(leftEye)
rightEyeHull = cv2.convexHull(rightEye)
#绘制轮廓
cv2.drawContours(img, [leftEyeHull], -1, (0, 0, 0), 1)
cv2.drawContours(img, [rightEyeHull], -1, (0, 0, 0), 1)
#取两只眼长宽比的的平均值作为每一帧的计算结果
eyeRatioVal = (leftEyesVal + rightEyesVal) / 2.0
# #嘴
# mouth=shape[leftMouth:rightMouth]
# mouthHull=cv2.convexHull(mouth)
# cv2.drawContours(frame, [mouthHull], -1, (0, 255, 0), 1)
# #鼻子
# nose=shape[leftNose:rightNose]
# noseHull=cv2.convexHull(nose)
# cv2.drawContours(frame, [noseHull], -1, (0, 255, 0), 1)
# #下巴
# jaw=shape[leftJaw:rightJaw]
# jawHull=cv2.convexHull(jaw)
# cv2.drawContours(frame, [jawHull], -1, (0, 255, 0), 1)
# #左眉毛
# leftEyebrow=shape[left_leftEyebrow:left_rightEyebrow]
# leftEyebrowHull=cv2.convexHull(leftEyebrow)
# cv2.drawContours(frame, [leftEyebrowHull], -1, (0, 255, 0), 1)
# #右眉毛
# rightEyebrow=shape[right_leftEyebrow:right_rightEyebrow]
# rightEyebrowHull=cv2.convexHull(rightEyebrow)
# cv2.drawContours(frame, [rightEyebrowHull], -1, (0, 255, 0), 1)
if collectCount < collectCountInterval:
collectCount+=1
collectSum+=eyeRatioVal
# cv2.putText(img, "DATA COLLECTING", (10, 10),cv2.FONT_HERSHEY_SIMPLEX, 0.2, (0, 0, 0), 1)
print("[INFO] DATA COLLECTING... ")
Adafruit_CharLCD.clear()
Adafruit_CharLCD.message("DATA COLLECTING")
startCheck=False
else:
if not startCheck:
eyesRatioLimit=collectSum/(1.0*collectCountInterval)
# Adafruit_CharLCD.clear()
Adafruit_CharLCD.clear()
Adafruit_CharLCD.message("fatigue detection")
print('眼睛长宽比均值',eyesRatioLimit)
# eyesRatioLimit = round(eyesRatioLimit,2)
# AverageEyesRatio = str(eyesRatioLimit)
startCheck=True
# Adafruit_CharLCD.clear()
# Adafruit_CharLCD.message("fatigue detection")
if startCheck:
#如果眼睛长宽比大于之前检测到的阈值，则计数，闭眼次数超过50次则认为已经"睡着"
if eyeRatioVal > 1.11*eyesRatioLimit:
eyesCloseCount += 1
if eyesCloseCount >= 22:
# cv2.putText(frame, "SLEEP!!!", (580, 30),cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 0, 255), 2)
# cv2.putText(img, "SLEEP!!!", (60, 40),cv2.FONT_HERSHEY_SIMPLEX, 0.2, (0, 0, 0), 1)
# if eyesCloseCount%2 == 0 :
#Adafruit_CharLCD.clear()
#Adafruit_CharLCD.message(str(eyesCloseCount)+'/n sleep!')
#time.sleep(0.1)
#Adafruit_CharLCD.message("sleep!")
self.buzzer.blink(0.3)
#time.sleep(0.5)
elif eyesCloseCount >= 4:
# cv2.putText(img, "EXHAUSTED!", (60, 40),cv2.FONT_HERSHEY_SIMPLEX, 0.2, (0, 0, 0), 1)
# if eyesCloseCount%2 == 0 :
#Adafruit_CharLCD.clear()
#Adafruit_CharLCD.message('close'+str(eyesCloseCount)+'/n exhausted!')
#time.sleep(0.1)
# Adafruit_CharLCD.message('close'+"exhausted!")
self.buzzer.blink(0.1)
#time.sleep(0.5)
elif eyesCloseCount >= 0:
# cv2.putText(img, "WIDE AWAKE", (60, 40),cv2.FONT_HERSHEY_SIMPLEX, 0.2, (0, 0, 0), 1)
#
#Adafruit_CharLCD.clear()
#Adafruit_CharLCD.message('close'+str(eyesCloseCount)+'/n Wake!')
#time.sleep(0.1)
print(datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S'))
#time.sleep(0.5)
print(datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S'))
# Adafruit_CharLCD.message("Wake!")
#print(datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S'))
#print(datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S'))
self.buzzer.off()
# alarm_beep(eyesCloseCount)
else:
eyesCloseCount = 0
print('眼睛实时长宽比:{:.2f} '.format(eyeRatioVal/eyesRatioLimit))
# eyeRatioVal = round(eyeRatioVal,2)
# Adafruit_CharLCD.clear()
#Adafruit_CharLCD.message('av:'+AverageEyesRatio+'\n now:'+str(eyeRatioVal))
#眼睛长宽比
# cv2.putText(img, "EYES_RATIO: {:.2f}".format(eyeRatioVal), (20, 20),cv2.FONT_HERSHEY_SIMPLEX, 0.2, (0, 0, 0), 1)
#闭眼次数
# cv2.putText(img,"EYES_COLSE: {}".format(eyesCloseCount),(40,30),cv2.FONT_HERSHEY_SIMPLEX,0.2,(0,0,0),1)
#通过检测嘴的长宽比检测有没有打哈欠，后来觉得没什么用
# cv2.putText(frame,"MOUTH_RATIO: {:.2f}".format(mouthRatio(mouth)),(500, 30),cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 255, 0), 2)
#cv2.imshow("Frame", img)
#key = cv2.waitKey(1) & 0xFF
#停止
#if key == ord("S"): break
if self.button_reset.is_pressed :
# if self.button_reset() :
print('program is over')
mixer.init()
mixer.music.load('xiaoge.mp3')
mixer.music.play()
break
cap.release()
def create_resource(ti=15,
v="Hybernation",
inac=5,
asleep=2,
fold="/home/satya/Pictures"):
# Initialize Pygame and load music
ft = 0
pygame.mixer.init()
pygame.mixer.music.load(r'assets\audio\faded.ogg')
# Minimum threshold of eye aspect ratio below which alarm is triggerd
EYE_ASPECT_RATIO_THRESHOLD = 0.3
# Minimum consecutive frames for which eye ratio is below threshold for alarm to be triggered
EYE_ASPECT_RATIO_CONSEC_FRAMES = 50
# COunts no. of consecutuve frames below threshold value
COUNTER = 0
b = 1
flag1 = True
global inace
# Load face cascade which will be used to draw a rectangle around detected faces.
face_cascade = cv2.CascadeClassifier(
r"assets\haarcascades\haarcascade_frontalface_default.xml")
# This function calculates and return eye aspect ratio
def eye_aspect_ratio(eye):
A = distance.euclidean(eye[1], eye[5])
B = distance.euclidean(eye[2], eye[4])
C = distance.euclidean(eye[0], eye[3])
# EAR = eye_aspect_ratio
ear = (A + B) / (2 * C)
return ear
# Load face detector and predictor, uses dlib shape predictor file
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(
r'assets\shape_predictor_68_face_landmarks.dat')
# Extract indexes of facial landmarks for the left and right eye
(lStart, lEnd) = face_utils.FACIAL_LANDMARKS_IDXS['left_eye']
(rStart, rEnd) = face_utils.FACIAL_LANDMARKS_IDXS['right_eye']
# Start webcam video capture /Give some time for camera to initialize(not required) /time.sleep(2)
a = 1
while (True):
video_capture = cv2.VideoCapture(0)
nowt = datetime.datetime.now()
nt1 = nowt.minute
newt = nowt + datetime.timedelta(minutes=ti)
nt2 = newt.minute
print("***********")
print(nowt, newt)
print("***********")
while (True):
# Read each frame and flip it, and convert to grayscale
ret, frame = video_capture.read()
frame = cv2.flip(frame, 1)
if ret:
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# Detect facial points through detector function
faces = detector(gray, 0)
# Detect faces through haarcascade_frontalface_default.xml
face_rectangle = face_cascade.detectMultiScale(gray, 1.3, 5)
# Draw rectangle around each face detected
for (x, y, w, h) in face_rectangle:
cv2.rectangle(frame, (x, y), (x + w, y + h), (255, 0, 0),
2)
# Detect facial points
# print(type(faces))
if len(faces) == 0:
COUNTER = 0
inacs = datetime.datetime.now()
# print(inacs)
if flag1:
inace = inacs + datetime.timedelta(minutes=inac)
inace = inace.minute
flag1 = False
if (inacs.minute > inace and b == 1 and flag1 == False):
pygame.mixer.music.play(-1)
b = 0
print(inacs, inace)
if (b == 0 and flag1 == False and inacs.minute >
(inace + 1)):
flag1 = True
b = 1
pygame.mixer.music.stop()
pic = pyautogui.screenshot()
sour = fold + "screenshot.png"
pic.save(sour)
ft = 1
for face in faces:
if (b != 1):
flag1 = True
b = 1
pygame.mixer.music.stop()
shape = predictor(gray, face)
shape = face_utils.shape_to_np(shape)
# Get array of coordinates of leftEye and rightEye
leftEye = shape[lStart:lEnd]
rightEye = shape[rStart:rEnd]
# Calculate aspect ratio of both eyes
leftEyeAspectRatio = eye_aspect_ratio(leftEye)
rightEyeAspectRatio = eye_aspect_ratio(rightEye)
eyeAspectRatio = (leftEyeAspectRatio +
rightEyeAspectRatio) / 2
# Use hull to remove convex contour discrepencies and draw eye shape around eyes
leftEyeHull = cv2.convexHull(leftEye)
rightEyeHull = cv2.convexHull(rightEye)
cv2.drawContours(frame, [leftEyeHull], -1, (0, 255, 0), 1)
cv2.drawContours(frame, [rightEyeHull], -1, (0, 255, 0), 1)
# Detect if eye aspect ratio is less than threshold
# print(eyeAspectRatio)
if (eyeAspectRatio < EYE_ASPECT_RATIO_THRESHOLD):
COUNTER += 1
# If no. of frames is greater than threshold frames,
print(COUNTER, EYE_ASPECT_RATIO_CONSEC_FRAMES)
if COUNTER >= EYE_ASPECT_RATIO_CONSEC_FRAMES and COUNTER >= asleep and a == 1:
pygame.mixer.music.play(-1)
a = 0
cv2.putText(frame, "You are Drowsy", (150, 200),
cv2.FONT_HERSHEY_SIMPLEX, 1.5,
(0, 0, 255), 2)
if COUNTER >= EYE_ASPECT_RATIO_CONSEC_FRAMES and COUNTER >= (
asleep + 20):
pic = pyautogui.screenshot()
sour = fold + "screenshot.png"
pic.save(sour)
pygame.mixer.music.stop()
ft = 1
break
else:
pygame.mixer.music.stop()
a = 1
COUNTER = 0
if ft == 1:
break
cv2.imshow('Video', frame)
kp = cv2.waitKey(1)
nt1 = datetime.datetime.now().minute
if (nt2 < nt1):
break
elif (kp & 0xFF == ord('q')):
return
if ft == 1:
video_capture.release()
cv2.destroyAllWindows()
pygame.mixer.music.stop()
break
else:
video_capture.release()
cv2.destroyAllWindows()
pygame.mixer.music.stop()
time.sleep(60 * ti)
if ft == 1:
time.sleep(10)
if v == "Hybernation":
changeStatus.hybernation()
else:
changeStatus.shut_down()
def main():
args = get_args()
model_name = args.model_name
weight_file = args.weight_file
margin = args.margin
image_dir = args.image_dir
if not weight_file:
weight_file = get_file(
"age_only_resnet50_weights.061-3.300-4.410.hdf5",
pretrained_model,
cache_subdir="pretrained_models",
file_hash=modhash,
cache_dir=Path(__file__).resolve().parent)
# for face detection
detector = dlib.get_frontal_face_detector()
# load model and weights
model = get_model(model_name=model_name)
model.load_weights(weight_file)
img_size = model.input.shape.as_list()[1]
image_generator = yield_images_from_dir(
image_dir) if image_dir else yield_images()
for img in image_generator:
input_img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
img_h, img_w, _ = np.shape(input_img)
# detect faces using dlib detector
detected = detector(input_img, 1)
faces = np.empty((len(detected), img_size, img_size, 3))
if len(detected) > 0:
for i, d in enumerate(detected):
x1, y1, x2, y2, w, h = d.left(), d.top(
), d.right() + 1, d.bottom() + 1, d.width(), d.height()
xw1 = max(int(x1 - margin * w), 0)
yw1 = max(int(y1 - margin * h), 0)
xw2 = min(int(x2 + margin * w), img_w - 1)
yw2 = min(int(y2 + margin * h), img_h - 1)
cv2.rectangle(img, (x1, y1), (x2, y2), (255, 0, 0), 2)
faces[i, :, :, :] = cv2.resize(
img[yw1:yw2 + 1, xw1:xw2 + 1, :], (img_size, img_size))
# predict ages and genders of the detected faces
results = model.predict(faces)
ages = np.arange(0, 101).reshape(101, 1)
predicted_ages = results.dot(ages).flatten()
# draw results
for i, d in enumerate(detected):
label = str(int(predicted_ages[i]))
draw_label(img, (d.left(), d.top()), label)
cv2.imshow("result", img)
key = cv2.waitKey(-1) if image_dir else cv2.waitKey(30)
if key == 27: # ESC
break
def main():
video_file = args.video_path
video_name = video_file.split("/")[-1]
cap = cv2.VideoCapture(video_file)
frame_count = 0
shape_predictor = dlib.shape_predictor(args.shape_predictor)
face_detector = dlib.get_frontal_face_detector()
transform = transforms.Compose(
[transforms.ToPILImage(),
transforms.ToTensor()])
model = extendNet(num_classes=8)
model = nn.DataParallel(model)
model = load_model(model_path=args.model_path, model=model)
model.eval()
prev_frame = None
prev_emotion = -1
continuous_emotion = 0
saved_faces = []
saved_frame_num = []
while cap.isOpened():
ret, frame = cap.read()
if ret is not True:
break
if frame is None:
continue
frame_count = frame_count + 1
if frame_count % frame_interval != 0:
continue
curr_frame = Frame(frame, face_detector, shape_predictor, frame_count)
if prev_frame is not None and curr_frame.face_rects is not None:
for curr_face_rect in curr_frame.face_rects:
if check_for_lip_movement(curr_frame, prev_frame,
curr_face_rect):
face_img = curr_frame.get_face_from_rect(curr_face_rect)
image_input_for_model = transform(
resize_face_image(face_img))
emotion_label = check_emotion_in_face(
image_input_for_model, model=model)
if emotion_label > 0:
if emotion_label == prev_emotion:
# extend the emotion scene for the face
continuous_emotion = continuous_emotion + 1
saved_faces.append(face_img)
saved_frame_num.append(frame_count)
if continuous_emotion >= streak_for_emotions:
folder_name = create_rename_folder(
saved_frame_num, emotion_label, video_name)
write_all_faces(saved_faces, saved_frame_num,
continuous_emotion,
emotion_label, folder_name,
video_name)
else:
# reinitialize the emotion and saved faces
continuous_emotion = 1
saved_faces = [face_img]
saved_frame_num = [frame_count]
prev_emotion = emotion_label
print(frame_count)
prev_frame = curr_frame
cap.release()
cv2.destroyAllWindows()
def main(args):
sleep(random.random())
output_dir = os.path.expanduser(args.output_dir)
if not os.path.exists(output_dir):
os.makedirs(output_dir)
# Store some git revision info in a text file in the log directory
src_path, _ = os.path.split(os.path.realpath(__file__))
facenet.store_revision_info(src_path, output_dir, ' '.join(sys.argv))
dataset = facenet.get_dataset(args.input_dir)
print('Loading predictor model')
# Create a HOG face detector using the built-in dlib class
face_detector = dlib.get_frontal_face_detector()
face_pose_predictor = dlib.shape_predictor(args.model)
face_aligner = openface.AlignDlib(args.model)
# Add a random key to the filename to allow alignment using multiple processes
random_key = np.random.randint(0, high=99999)
nrof_images_total = 0
nrof_successfully_aligned = 0
nrof_unchanged_images = 0
print('Total number of classes:', len(dataset))
print('Number of classes to align:', args.num_classes)
for cls in dataset[:args.num_classes]:
print('Processing class:', cls.name)
output_class_dir = os.path.join(output_dir, cls.name)
if not os.path.exists(output_class_dir):
os.makedirs(output_class_dir)
for image_path in cls.image_paths:
nrof_images_total += 1
filename = os.path.splitext(os.path.split(image_path)[1])[0]
output_filename = os.path.join(output_class_dir, filename + '.png')
if not os.path.exists(output_filename):
try:
img = misc.imread(image_path)
except (IOError, ValueError, IndexError) as e:
errorMessage = '{}: {}'.format(image_path, e)
print(errorMessage)
else:
if img.ndim < 2:
print('Unable to align "%s"' % image_path)
continue
if img.ndim == 2:
img = facenet.to_rgb(img)
img = img[:, :, 0:3]
# Run the HOG face detector on the image data
detected_face = face_detector(img, 1)
if len(detected_face) > 0:
# As face(s) detected is/ are stored in a list, we need to pull it out.
face_rect = detected_face[0]
# Get the the face's pose
pose_landmarks = face_pose_predictor(img, face_rect)
# Use openface to calculate and perform the face alignment
alignedFace = face_aligner.align(
534,
img,
face_rect,
landmarkIndices=openface.AlignDlib.
OUTER_EYES_AND_NOSE)
misc.imsave(output_filename, alignedFace)
nrof_successfully_aligned += 1
else:
filename_base, file_extension = os.path.splitext(
output_filename)
# Add 'fail' to indicates that dlib cannot detect face in this image.
output_filename = "{}{}".format(
filename_base + '_failed', file_extension)
nrof_unchanged_images += 1
misc.imsave(output_filename, img)
print('Total number of images:', nrof_images_total)
print('Number of successfully aligned images:', nrof_successfully_aligned)
print(
'Number of resized images (MTCNN cannot detect face in these images):',
nrof_unchanged_images)
def get_mouse_dlib(video_path,
save_path,
frame_detection=float("inf"),
save_pic_or_not=None):
"""
:param video_path: 视频路径
:param save_path: 存储路径
:param frame_detection: 处理多少帧，默认所有
:param save_pic_or_not: 是否保存标记出68点图片
:return: None
"""
video_name = os.path.basename(video_path)[:-4]
save_path = os.path.join(save_path, video_name)
if not os.path.exists(save_path):
os.makedirs(save_path)
save_picture = save_path + "/picture"
save_mouse_picutre = save_path + "/mouse_picture"
if not os.path.exists(save_picture):
os.makedirs(save_picture)
if not os.path.exists(save_mouse_picutre):
os.makedirs(save_mouse_picutre)
save_txt = save_path + "/txt"
if not os.path.exists(save_txt):
os.makedirs(save_txt)
print("make_path")
x_txt = open(save_txt + "/x.txt", "a+") #记录刻画嘴部的20个点的横坐标，以及脸宽
y_txt = open(save_txt + "/y.txt", "+a") #记录刻画嘴部的20个点的纵坐标，以及脸宽
cap = cv2.VideoCapture(video_path)
frame_index = 0
zero = 0
data_ = []
all_data = {}
if cap.isOpened():
success = True
else:
success = False
print("读取失败!")
return None
while (success and frame_detection > 0):
frame_detection = frame_detection - 1
success, frame = cap.read()
if not success:
break
frame = np.rot90(frame, -1)
frame = np.rot90(frame, -1)
""" """
detector = dlib.get_frontal_face_detector()
# dlib的68点模型，使用作者训练好的特征预测器
predictor = dlib.shape_predictor(
r"C:\Users\admin\Desktop\net_for_mouse\project_for_mouse\shape_predictor_68_face_landmarks.dat"
)
# 特征提取器的实例化
dets = detector(frame, 1)
print("人脸数：", len(dets))
if len(dets) == 0:
print("未检测到人脸")
cv2.imwrite(save_picture + "/{}wrong.jpg".format(zero), frame)
zero += 1
continue
elif (len(dets) == 1):
frame_index += 1
for k, d in enumerate(dets):
print("第", frame_index, "个人脸")
width = d.right() - d.left()
heigth = d.bottom() - d.top()
shape = predictor(frame, d)
print('人脸面积为：', (width * heigth))
if save_pic_or_not:
# frame = frame[..., ::-1]
img = cv2.cvtColor(frame, cv2.COLOR_RGB2GRAY)
img1 = img[
int(max(shape.part(49).y,
shape.part(50).y) - 5
):int(min(shape.part(57).y,
shape.part(58).y) + 5),
int(shape.part(48).x - 5):int(shape.part(54).x + 5)]
cv2.imwrite(
save_mouse_picutre + "/{}.jpg".format(frame_index),
img1)
for i in range(68):
frame = cv2.circle(frame,
(shape.part(i).x, shape.part(i).y),
2, (0, 255, 0), 1, 3)
cv2.imwrite(save_picture + "/{}.jpg".format(frame_index),
frame)
pose = []
score = [1] * 20
skeleton = {}
for j in range(48, 68):
x_txt.write(str(shape.part(j).x) + " ")
pose.append(shape.part(j).x)
pose.append(shape.part(j).y)
x_txt.write(str(width))
x_txt.write("\n")
for j in range(48, 68):
y_txt.write(str(shape.part(j).y) + " ")
y_txt.write(str(heigth))
y_txt.write("\n")
skeleton["pose"] = pose
skeleton["score"] = score
skeleton_ = []
skeleton_.append(skeleton)
data = {}
data["frame_index"] = frame_index
data["skeleton"] = skeleton_
data_.append(data)
all_data["data"] = data_
all_data["label"] = str(video_path.split("/")[-2])
all_data["label_index"] = int(video_path.split("/")[-2])
with open(
save_path + "/" + str(video_path.split("/")[-2]) + "-" +
str(video_path.split("/")[-1][:-4]) + ".json", "w") as f:
json.dump(all_data, f)
class FaceAction:
tot = 0
detect = get_frontal_face_detector()
predict = shape_predictor("shape_predictor_68_face_landmarks.dat")
(lStart, lEnd) = face_utils.FACIAL_LANDMARKS_68_IDXS["left_eye"]
(rStart, rEnd) = face_utils.FACIAL_LANDMARKS_68_IDXS["right_eye"]
(mStart, mEnd) = face_utils.FACIAL_LANDMARKS_68_IDXS["mouth"]
K = [
6.5308391993466671e+002, 0.0, 3.1950000000000000e+002, 0.0,
6.5308391993466671e+002, 2.3950000000000000e+002, 0.0, 0.0, 1.0
]
D = [
7.0834633684407095e-002, 6.9140193737175351e-002, 0.0, 0.0,
-1.3073460323689292e+000
]
cam_matrix = np.array(K).reshape(3, 3).astype(np.float32)
dist_coeffs = np.array(D).reshape(5, 1).astype(np.float32)
object_pts = np.float32([[6.825897, 6.760612, 4.402142],
[1.330353, 7.122144, 6.903745],
[-1.330353, 7.122144, 6.903745],
[-6.825897, 6.760612, 4.402142],
[5.311432, 5.485328, 3.987654],
[1.789930, 5.393625, 4.413414],
[-1.789930, 5.393625, 4.413414],
[-5.311432, 5.485328, 3.987654],
[2.005628, 1.409845, 6.165652],
[-2.005628, 1.409845, 6.165652],
[2.774015, -2.080775, 5.048531],
[-2.774015, -2.080775, 5.048531],
[0.000000, -3.116408, 6.097667],
[0.000000, -7.415691, 4.070434]])
reprojectsrc = np.float32([[10.0, 10.0, 10.0], [10.0, 10.0, -10.0],
[10.0, -10.0, -10.0], [10.0, -10.0, 10.0],
[-10.0, 10.0, 10.0], [-10.0, 10.0, -10.0],
[-10.0, -10.0, -10.0], [-10.0, -10.0, 10.0]])
def eye_aspect_ratio(self, eye):
A = distance.euclidean(eye[1], eye[5])
B = distance.euclidean(eye[2], eye[4])
C = distance.euclidean(eye[0], eye[3])
ear = (A + B) / (2.0 * C)
return ear
def mouth_aspect_ratio(self, mouth):
A = distance.euclidean(mouth[13], mouth[19])
B = distance.euclidean(mouth[14], mouth[18])
C = distance.euclidean(mouth[15], mouth[17])
D = distance.euclidean(mouth[12], mouth[16])
mar = (A + B + C) / (2.0 * D)
return mar
def drowsy(self, frame):
frame = resize(frame, width=450)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
subjects = self.detect(gray, 0)
self.tot = len(subjects)
# print(len(subjects))
# print(self.tot)
if (len(subjects) == 0):
return 1
for subject in subjects:
shape = self.predict(gray, subject)
shape = face_utils.shape_to_np(shape)
leftEye = shape[self.lStart:self.lEnd]
rightEye = shape[self.rStart:self.rEnd]
leftEAR = self.eye_aspect_ratio(leftEye)
rightEAR = self.eye_aspect_ratio(rightEye)
ear = (leftEAR + rightEAR) / 2.0
return ear
def yawn(self, frame):
frame = resize(frame, width=450)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
subjects = self.detect(gray, 0)
if (len(subjects) == 0):
return 0
for subject in subjects:
shape = self.predict(gray, subject)
shape = face_utils.shape_to_np(shape)
mouth = shape[self.mStart:self.mEnd]
mar = self.mouth_aspect_ratio(mouth)
return mar
def get_head_pose(self, shape, object_pts, cam_matrix, dist_coeffs,
reprojectsrc):
image_pts = np.float32([
shape[17], shape[21], shape[22], shape[26], shape[36], shape[39],
shape[42], shape[45], shape[31], shape[35], shape[48], shape[54],
shape[57], shape[8]
])
_, rotation_vec, translation_vec = cv2.solvePnP(
object_pts, image_pts, cam_matrix, dist_coeffs)
reprojectdst, _ = cv2.projectPoints(reprojectsrc, rotation_vec,
translation_vec, cam_matrix,
dist_coeffs)
reprojectdst = tuple(map(tuple, reprojectdst.reshape(8, 2)))
# calc euler angle
rotation_mat, _ = cv2.Rodrigues(rotation_vec)
pose_mat = cv2.hconcat((rotation_mat, translation_vec))
_, _, _, _, _, _, euler_angle = cv2.decomposeProjectionMatrix(pose_mat)
return reprojectdst, euler_angle
def head_pose(self, frame):
face_rects = self.detect(frame, 0)
if (len(face_rects) > 0):
shape = self.predict(frame, face_rects[0])
shape = face_utils.shape_to_np(shape)
_, euler_angle = self.get_head_pose(shape, self.object_pts,
self.cam_matrix,
self.dist_coeffs,
self.reprojectsrc)
if (-10 <= euler_angle[2, 0] and euler_angle[2, 0] <= 10):
return 0
else:
return 1
else:
return 1
def run_frame(self, frame):
return (self.drowsy(frame), self.yawn(frame), self.head_pose(frame),
self.tot)
def detect_blink():
# construct the argument parse and parse the arguments
# ap = argparse.ArgumentParser()
# ap.add_argument("-p", "--shape-predictor", required=True,
# help="path to facial landmark predictor")
# ap.add_argument("-v", "--video", help="path to input video file")
# args = vars(ap.parse_args())
# define two constants, one for the eye aspect ratio to indicate
# blink and then a second constant for the number of consecutive
# frames the eye must be below the threshold
EYE_AR_THRESH = 0.23
EYE_AR_CONSEC_FRAMES = 3
# initialize the frame counters and the total number of blinks
COUNTER = 0
TOTAL = 0
# initialize dlib's face detector (HOG-based) and then create
# the facial landmark predictor
print("[INFO] loading facial landmark predictor...")
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(
"C:\\Users\\t-amsi\\Eye-blink-detection\\WFHealth\\Modules\\shape_predictor_68_face_landmarks.dat"
)
# grab the indexes of the facial landmarks for the left and
# right eye, respectively
(lStart, lEnd) = face_utils.FACIAL_LANDMARKS_IDXS["left_eye"]
(rStart, rEnd) = face_utils.FACIAL_LANDMARKS_IDXS["right_eye"]
# start the video stream thread
print("[INFO] starting video stream thread...")
# vs = FileVideoStream(args["video"]).start()
# fileStream = True
# vs = VideoStream(src=0).start()
vs = cv2.VideoCapture('../blink.mp4')
# vs = VideoStream(usePiCamera=True).start()
fileStream = False
time.sleep(1.0)
# loop over frames from the video stream
while vs.isOpened():
# if this is a file video stream, then we need to check if
# there any more frames left in the buffer to process
# if fileStream and not vs.more():
# break
# grab the frame from the threaded video file stream, resize
# it, and convert it to grayscale
# channels)
# frame = vs.read()
ret, frame = vs.read()
if ret == False:
break
# gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# cv2.imshow('frame',gray)
# if cv2.waitKey(1) & 0xFF == ord('q'):
# breakcap.release()
# cv2.destroyAllWindows()
frame = imutils.resize(frame, width=450)
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# detect faces in the grayscale frame
rects = detector(gray, 0)
# loop over the face detections
for rect in rects:
# determine the facial landmarks for the face region, then
# convert the facial landmark (x, y)-coordinates to a NumPy
# array
shape = predictor(gray, rect)
shape = face_utils.shape_to_np(shape)
# extract the left and right eye coordinates, then use the
# coordinates to compute the eye aspect ratio for both eyes
leftEye = shape[lStart:lEnd]
rightEye = shape[rStart:rEnd]
leftEAR = eye_aspect_ratio(leftEye)
rightEAR = eye_aspect_ratio(rightEye)
# average the eye aspect ratio together for both eyes
ear = (leftEAR + rightEAR) / 2.0
# compute the convex hull for the left and right eye, then
# visualize each of the eyes
leftEyeHull = cv2.convexHull(leftEye)
rightEyeHull = cv2.convexHull(rightEye)
cv2.drawContours(frame, [leftEyeHull], -1, (0, 255, 0), 1)
cv2.drawContours(frame, [rightEyeHull], -1, (0, 255, 0), 1)
# check to see if the eye aspect ratio is below the blink
# threshold, and if so, increment the blink frame counter
if ear < EYE_AR_THRESH:
COUNTER += 1
# otherwise, the eye aspect ratio is not below the blink
# threshold
else:
# if the eyes were closed for a sufficient number of
# then increment the total number of blinks
if COUNTER >= EYE_AR_CONSEC_FRAMES:
TOTAL += 1
# reset the eye frame counter
COUNTER = 0
# draw the total number of blinks on the frame along with
# the computed eye aspect ratio for the frame
# cv2.putText(frame, "Blinks: {}".format(TOTAL), (10, 30),
# cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
# cv2.putText(frame, "EAR: {:.2f}".format(ear), (300, 30),
# cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
# # show the frame
# cv2.imshow("Frame", frame)
# key = cv2.waitKey(1) & 0xFF
# # if the `q` key was pressed, break from the loop
# if key == ord("q"):
# break
# do a bit of cleanup
cv2.destroyAllWindows()
vs.release()
# print(COUNTER)
return COUNTER
def startRec():
depth = 16
k = 8
margin = 0.4
image_dir = nameEdit.text()
weight_file = get_file("weights.28-3.73.hdf5",
pretrained_model,
cache_subdir="pretrained_models",
file_hash=modhash,
cache_dir=str(Path(__file__).resolve().parent))
# for face detection
detector = dlib.get_frontal_face_detector()
# load model and weights
img_size = 64
model = WideResNet(img_size, depth=depth, k=k)()
model.load_weights(weight_file)
image_generator = yield_images_from_dir(
image_dir) if image_dir else yield_images()
count_im_in_dir = count_im_dir(image_dir)
count = 0
for img in image_generator:
#print(count)
input_img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
img_h, img_w, _ = np.shape(input_img)
# detect faces using dlib detector
detected = detector(input_img, 1)
faces = np.empty((len(detected), img_size, img_size, 3))
if len(detected) > 0:
for i, d in enumerate(detected):
x1, y1, x2, y2, w, h = d.left(), d.top(
), d.right() + 1, d.bottom() + 1, d.width(), d.height()
xw1 = max(int(x1 - margin * w), 0)
yw1 = max(int(y1 - margin * h), 0)
xw2 = min(int(x2 + margin * w), img_w - 1)
yw2 = min(int(y2 + margin * h), img_h - 1)
cv2.rectangle(img, (x1, y1), (x2, y2), (255, 0, 0), 2)
# cv2.rectangle(img, (xw1, yw1), (xw2, yw2), (255, 0, 0), 2)
faces[i, :, :, :] = cv2.resize(
img[yw1:yw2 + 1, xw1:xw2 + 1, :], (img_size, img_size))
# predict ages and genders of the detected faces
results = model.predict(faces)
predicted_genders = results[0]
ages = np.arange(0, 101).reshape(101, 1)
predicted_ages = results[1].dot(ages).flatten()
# draw results
for i, d in enumerate(detected):
label = "{}, {}".format(
int(predicted_ages[i]),
"M" if predicted_genders[i][0] < 0.5 else "F")
draw_label(img, (d.left(), d.top()), label)
#cv2.imshow("result", img)
cv2.imwrite(name1Edit.text() + "/res" + str(count) + ".jpg", img)
count += 1
progBarUpdate(100 * (count / count_im_in_dir))
key = cv2.waitKey(-1) if image_dir else cv2.waitKey(30)
if key == 27: # ESC
break
cv2.destroyAllWindows()
def run(self, bpm):
# Capture video and set up dlib.
cap = cv2.VideoCapture(0)
hog_face_detector = dlib.get_frontal_face_detector()
dlib_facelandmark = dlib.shape_predictor(
"data_files/shape_predictor_68_face_landmarks.dat")
self.keep_recording = True
while self.keep_recording:
# Determine blinks per min
cur_time = time.time()
#blink_queue = [x for x in self.blink_queue if cur_time - x < 60]
#self.bpm = len(self.blink_queue)
self.bpm = round(self.total_bpm * 60 /
(cur_time - self.start_time))
bpm = self.bpm
drowsy_window = [
x for x in self.drowsy_window if cur_time - x < 10
]
if len(drowsy_window) >= 5:
self.condition = 'bad'
elif self.bpm < 15:
self.condition = 'bad'
else:
self.condition = 'good'
#if cur_time - start_time >= 60:
# blinks_pm = cur_bpm
# cur_bpm = 0
# start_time = cur_time
_, frame = cap.read()
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
faces = hog_face_detector(gray)
for face in faces:
face_landmarks = dlib_facelandmark(gray, face)
leftEye = []
rightEye = []
for n in range(36, 42):
x = face_landmarks.part(n).x
y = face_landmarks.part(n).y
leftEye.append((x, y))
next_point = n + 1
if n == 41:
next_point = 36
x2 = face_landmarks.part(next_point).x
y2 = face_landmarks.part(next_point).y
cv2.line(frame, (x, y), (x2, y2), (0, 255, 0), 1)
for n in range(42, 48):
x = face_landmarks.part(n).x
y = face_landmarks.part(n).y
rightEye.append((x, y))
next_point = n + 1
if n == 47:
next_point = 42
x2 = face_landmarks.part(next_point).x
y2 = face_landmarks.part(next_point).y
cv2.line(frame, (x, y), (x2, y2), (0, 255, 0), 1)
#print(leftEye)
#print(rightEye)
left_ear = calculate_EAR(leftEye)
right_ear = calculate_EAR(rightEye)
# EAR = EYE ASPECT RATIO
EAR = (left_ear + right_ear) / 2
EAR = round(EAR, 2)
# Default threshold for drowsiness
if EAR < 0.26:
cv2.putText(frame, "DROWSY", (20, 100),
cv2.FONT_HERSHEY_SIMPLEX, 2, (0, 0, 255), 4)
#cv2.putText(frame,"Are you Sleepy?",(20,400), cv2.FONT_HERSHEY_SIMPLEX,2,(0,0,255),4)
#print("Drowsy")
if self.drowsy_window and (round(self.drowsy_window[-1]) !=
round(cur_time)):
self.drowsy_window.append(cur_time)
elif not self.drowsy_window:
self.drowsy_window.append(cur_time)
# Threshold for blinking
if EAR < 0.2 and not self.eye_closed:
self.total_bpm += 1
self.blink_queue.append(cur_time)
self.eye_closed = True
elif EAR >= 0.2:
self.eye_closed = False
#print(EAR)
# Display GUI frame
cv2.putText(frame, f'BPM: {self.bpm}', (360, 50),
cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 4)
cv2.putText(frame, f'Total BPM: {self.total_bpm}', (360, 100),
cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 4)
cv2.putText(frame, f'Condition: {self.condition}', (180, 350),
cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 4)
cv2.imshow("Are you Sleepy", frame)
key = cv2.waitKey(1)
if key == 27:
break
cap.release()
cv2.destroyAllWindows()
def get_detector():
"""
:return: A face detector.
"""
return dlib.get_frontal_face_detector()
def main():
detector=dlib.get_frontal_face_detector()
eyes_calibrated=False
total_frames_count=0
eye_contact_frames=0
predictor=dlib.shape_predictor("shape_predictor_68_face_landmarks.dat")
lefteye_vertical_line=0
lefteye_horizontal_line=0
righteye_vertical_line=0
righteye_horizontal_line=0
cap=cv2.VideoCapture('video2.mp4')
while True:
if total_frames_count%15==0:
ret, frame = cap.read()
if ret and frame is not None:
frame = cv2.rotate(frame, cv2.ROTATE_90_CLOCKWISE)
#frame=increase_contrast(frame,cv2) --Don't Apply Contrast on the image because the contours won't be detected
frame=cv2.cvtColor(frame,cv2.COLOR_BGR2GRAY)
faces=detector(frame)
shape=predictor(frame,faces[0])
#Left Eye
lefteye=extract_left_eye(shape,frame)
_,lefteye=cv2.threshold(lefteye,0,255,cv2.THRESH_BINARY_INV)
lefteye_contour=return_biggest_contour_in_image(lefteye)
if lefteye_contour is None:
continue
#Right Eye
righteye=extract_right_eye(shape,frame)
_,righteye=cv2.threshold(righteye,0,255,cv2.THRESH_BINARY_INV)
righteye_contour=return_biggest_contour_in_image(righteye)
if righteye_contour is None:
continue
#calibrate eyes for finding the center poin
left_v,left_h=detect_contour_location(lefteye_contour,cv2)
right_v,right_h=detect_contour_location(righteye_contour,cv2)
if eyes_calibrated is False:
print("Setting the eye midpoints")
lefteye_vertical_line,lefteye_horizontal_line=left_v,left_h
righteye_vertical_line,righteye_horizontal_line=right_v,right_h
print("Calibrated: left (%f, %f) Right(%f, %f) " % (left_v,left_h,right_v,right_h))
eyes_calibrated=True
else:
#determine lefteyeContact
is_left_eye_contact_maintained=detect_eye_contact(left_v,left_h,lefteye_vertical_line,lefteye_horizontal_line)
is_right_eye_contact_maintained=detect_eye_contact(right_v,right_h,righteye_vertical_line,righteye_horizontal_line)
if is_left_eye_contact_maintained and is_right_eye_contact_maintained:
eye_contact_frames=eye_contact_frames+15
else:
#path='C:\Hardik Personal Data\C Drive Data\EyeContactDetection\eyeContactGone\image{}.png'.format(count)
#cv2.imwrite(path,frame)
print("Eye Contact is gone: left (%f, %f) Right(%f, %f) " % (left_v,left_h,right_v,right_h))
cv2.imshow('Video',frame)
cv2.imshow('Left Eye',lefteye)
cv2.imshow('Right Eye',righteye)
if cv2.waitKey(1) & 0xFF == ord('q'):
cap.release()
cv2.destroyAllWindows()
break
else:
break
total_frames_count=total_frames_count+1
print("Final Result:")
print("Percentage of eyecontact maintained is:")
print((eye_contact_frames/total_frames_count)*100)
cap.release()
cv2.destroyAllWindows()
