def predict():
# this will contain the
data = {"success": False}
# for keeping track of authentication status
data['authenticate'] = False
# ensure an image was properly uploaded to our endpoint
if flask.request.method == "POST":
if flask.request.files.get("image"):
# read the image in PIL format
image = flask.request.files["image"].read()
image = np.array(Image.open(io.BytesIO(image)))
# save the image on server side
cv2.imwrite('saved_image/new.jpg', cv2.cvtColor(image, cv2.COLOR_RGB2BGR))
# CHECK FOR FACE IN THE IMAGE
valid_face = False
valid_face = face_present('saved_image/new.jpg')
# do facial recognition only when there is a face inside the frame
if valid_face:
# find image encoding and see if the image is of a registered user or not
encoding = img_to_encoding('saved_image/new.jpg', model)
min_dist, identity, authenticate = face_recognition(
encoding, user_db, model, threshold=0.9)
# save the output for sending as json
data["min_dist"] = str(min_dist)
data['email'] = identity
if identity != 'Unknown Person':
data['name'] = user_db[identity]['name']
else:
data['name'] = 'Unknown Person'
data['face_present'] = True
data['authenticate'] = authenticate
else:
# save the output for sending as json
data["min_dist"] = 'NaN'
data['identity'] = 'NaN'
data['name'] = 'NaN'
data['face_present'] = False
data['authenticate'] = False
print('No subject detected !')
# indicate that the request was a success
data["success"] = True
# create a new session
Session = sessionmaker(bind=engine)
s = Session()
# check if the user is logged in
if data['authenticate']:
session['logged_in'] = True
else:
flash('Unknown Person!')
# return the data dictionary as a JSON response
return flask.jsonify(data)
def add_user_img_path(user_db, FRmodel, name, img_path):
if name not in user_db:
user_db[name] = img_to_encoding(img_path, FRmodel)
# save the database
with open(r'C:\Users\KIIT\Desktop\Deep Learning\Face Detection based Attendence\database\user_dict.pickle', 'wb') as handle:
pickle.dump(user_db, handle, protocol=pickle.HIGHEST_PROTOCOL)
res='User ' + name + ' added successfully'
message.configure(text=res)
else:
res='The name is already registered! Try a different name.........'
message.configure(text=res)
def add_face():
data = {"face_present": False}
encoding = None
# CHECK FOR FACE IN THE IMAGE
valid_face = False
valid_face = face_present('saved_image/new.jpg')
# add user only if there is a face inside the picture
if valid_face:
# create image encoding
encoding = img_to_encoding('saved_image/new.jpg', model)
# save the output for sending as json
data['face_present'] = True
else:
# save the output for sending as json
data['face_present'] = False
print('No subject detected !')
return data, encoding
def find_face_realtime(image_path, database, model, threshold):
# find the face encodings for the input image
encoding = img_to_encoding(image_path, model)
registered = False
min_dist = 100
identity = 'Unknown Person'
# loop over all the recorded encodings in database
for name in database:
# find the similarity between the input encodings and claimed person's encodings using L2 norm
dist = np.linalg.norm(np.subtract(database[name], encoding))
print('distance for %s is %s' % (name, dist))
# check if minimum distance or not
if dist < min_dist:
min_dist = dist
identity = name
if min_dist > threshold:
registered = False
else:
registered = True
return min_dist, identity, registered