def init(kfn, kfe):
#temp vars : names, flag, floc, fe, ctr, prev_name, video, width, height, vname, out,
# rframe, rgbrframe, matched_faces, fdist, best_match, name, gray,
names = []
flag = True
floc = []
fe = []
ctr=0
name = " "
prev_name = " "
print('Press " e " | " r " to enroll..')
video = cv2.VideoCapture(0) #start cam
width = int(video.get(3))
height = int(video.get(4))
vname = 'output.avi'
out = cv2.VideoWriter(vname,cv2.VideoWriter_fourcc('M','J','P','G'), 10, (width,height))
while(1):
ret,frame = video.read() #grab frame by frame while(1)
rframe = cv2.resize(frame,(0,0),fx=0.25,fy=0.25) #not needed ,
#just to make the process faster
rgbrframe = cv2.cvtColor(rframe,cv2.COLOR_BGR2RGB)#cv2 uses BGR color whereas,
#face_rec uses RGB , so reverse content
out.write(frame) # write to videoLog
if flag:
floc = face_rec.face_locations(rgbrframe) # grab face from frame
fe = face_rec.face_encodings(rgbrframe,floc) # grab face encodings from frame
for fenc in fe:
matched_faces = face_rec.compare_faces(kfe,fenc)
fdist = face_rec.face_distance(kfe,fenc)
best_match = np.argmin(fdist)
if matched_faces[best_match]:
try:
name = kfn[best_match]
if prev_name != name:
print(name + ' - '+str(datetime.datetime.now()))
prev_name = name
except Exception:
pass
else:
name = 'Unknown'
if prev_name != name:
print('\t!!! Security Alert !!!\n\t\tDetected ' + name + '- ' + str(datetime.datetime.now()))
prev_name = name
gray = cv2.cvtColor(rframe,cv2.COLOR_BGR2GRAY)#converting unknowface frame to gray
faceCascade = cv2.CascadeClassifier(cv2.data.haarcascades + "haarcascade_frontalface_default.xml")#haarcascade unknown person front face
faces = faceCascade.detectMultiScale(gray,
scaleFactor=1.3,
minNeighbors=3,
minSize=(50, 50)
)
if len(faces):#faces contails multiple unknown faces in single frame
for (x, y, w, h) in faces:
roi_color = rframe[y : y + 240 , x : x + 360] #cropping unknown face
ctr = ctr + 1
path = 'unknownDetected/'
cv2.imwrite(os.path.join(path , str(ctr)+'Unknown_Face.jpg'), roi_color) #Saving cropped image of unknown to unknownDetected Folder
filename = path + str(ctr)+'Unknown_Face.jpg'
f1 = face_rec.load_image_file(filename)
try:
f1_encoding = face_rec.face_encodings(f1)[0]#Encoding unknown face
except: pass
kfe.append(f1_encoding)
s = 'Suspect'+str(ctr) #Detected Unknown face encoded as Suspect :)
kfn.append(s)
r1 = r.randint(0,100):
if r1 / 2 == 0:
s.sframes(name,frame)
flag = not flag
names += [name]
cv2.imshow('Video', frame) #show frames as being processed.
print('Here',*names)
#No markings are done on live frames so as to reduce the flickering on screen/video feed. Instead, as like on generic systems, results are shown on a terminal/console.
#TODO : Show on Text Area while adding Flask front-end.
if cv2.waitKey(1) & 0xFF == ord('q'):
out.release()
video.release()
cv2.destroyAllWindows()
break
if cv2.waitKey(1) & 0XFF == ord('e') or cv2.waitKey(1) & 0XFF == ord('r'):
print('Begin Enrollment....')
video.release()
cv2.destroyAllWindows()
e.enroll()
s.store(names)
init()
return
rframe = cv2.resize(frame,(0,0),fx=0.25,fy=0.25) #not needed ,
#just to make the process faster
rgbrframe = cv2.cvtColor(rframe,cv2.COLOR_BGR2RGB)#cv2 uses BGR color whereas,
#face_rec uses RGB , so reverse content
if flag:
floc = face_rec.face_locations(rgbrframe) # grab face from frame
fe = face_rec.face_encodings(rgbrframe,floc) # grab face encodings from frame
names = []
for fenc in fe:
matched_faces = face_rec.compare_faces(kfe,fenc)
name = 'Unknown'
fdist = face_rec.face_distance(kfe,fenc)
best_match = np.argmin(fdist)
if matched_faces[best_match]:
name = kfn[best_match]
names.append(name)
flag = not flag
# Display the results
for (top, right, bottom, left), name in zip(floc, names):
top *= 4 # resize image back again by *0.25
right *= 4
bottom *= 4
left *= 4
cv2.rectangle(frame, (left, top), (right, bottom), (0, 255, 0), 2)# Draw a box around the face
def findout(image, iname):
print(image)
url = 'rtsp://192.168.43.1:8080/h264_ulaw.sdp'
# url = 'http://192.168.43.1:8080/video'
video = cv2.VideoCapture(url)
#loading sample pictures
f1 = face_rec.load_image_file(image)
#learn how to recognise it
f1_encoding = face_rec.face_encodings(f1)[0]
#array for known encodings
kfe = [f1_encoding]
#array for known face names
kfn = [iname]
names = []
flag = True
floc = []
fe = []
count = 0
while (1):
ret, frame = video.read() #grab frame by frame while(1)
rframe = cv2.resize(frame, (0, 0), fx=0.25, fy=0.25) #not needed ,
#just to make the process faster
rgbrframe = cv2.cvtColor(
rframe, cv2.COLOR_BGR2RGB) #cv2 uses BGR color whereas,
#face_rec uses RGB , so reverse content
if flag:
floc = face_rec.face_locations(rgbrframe) # grab face from frame
fe = face_rec.face_encodings(
rgbrframe, floc) # grab face encodings from frame
names = []
for fenc in fe:
matched_faces = face_rec.compare_faces(kfe, fenc)
name = 'Unknown'
fdist = face_rec.face_distance(kfe, fenc)
best_match = np.argmin(fdist)
if matched_faces[best_match]:
name = kfn[best_match]
names.append(name)
flag = not flag
# Display the results
for (top, right, bottom, left), name in zip(floc, names):
top *= 4 # resize image back again by *0.25
right *= 4
bottom *= 4
left *= 4
cv2.rectangle(frame, (left, top), (right, bottom), (0, 255, 0),
4) # Draw a box around the face
cv2.rectangle(frame, (left, bottom - 35), (right, bottom),
(0, 0, 0), cv2.FILLED)
font = cv2.FONT_HERSHEY_DUPLEX
cv2.putText(frame, name, (left + 6, bottom - 6), font, 1.0,
(255, 255, 255), 1) #label the face
h, w, l = frame.shape
new_h = int(h / 2)
new_w = int(w / 2)
rzframe = cv2.resize(frame, (new_w, new_h))
cv2.imshow('Cam_feed', rzframe)
count += 1
if cv2.waitKey(1) & 0xFF == ord('q'):
print('processed ', count, 'frames')
video.release()
cv2.destroyAllWindows()
break