def classify(self, path_to_img):
# img = cv2.imread(path_to_img,0) # read as a grayscale image
img = detect_face(path_to_img)
if (img is None):
return -2
img_col = np.array(img, dtype='float64').flatten() # flatten the image
img_col -= self.mean_img_col # subract the mean column
img_col = np.reshape(img_col,
(self.mn, 1)) # from row vector to col vector
S = self.evectors.transpose(
) * img_col # projecting the normalized probe onto the
# Eigenspace, to find out the weights
diff = self.W - S # finding the min ||W_j - S||
norms = np.linalg.norm(diff, axis=0)
print(np.min(norms))
closest_face_id = np.argmin(
norms) # the id [0..240) of the minerror face to the sample
#threshold for not detecting any face
if np.min(norms) >= 6500:
return -1
return self.names[(
closest_face_id //
self.train_faces_count)] # return the faceid (1..40)
def registration_controller():
error = 0;
outer_id = '12345678helloAi'
f = open("/home/pi/Desktop/ProfMate/student_info.txt", mode='r')
info = f.readlines()
picture_addr = info[0].strip()
first_name = info[1].strip()
last_name = info[2].strip()
student_id = int(info[3].strip())
print(first_name)
print(last_name)
#if multiple, consider for loop
student_courses = 101
face_tokens = detect.detect_face(picture_addr)
if(len(face_tokens) > 1):
print("Invalid Picture: Multiple Face Detected")
#terminate
error = 1
return error
face_token = face_tokens[0]
newStudent = studentControl(first_name, last_name, student_id, student_courses, outer_id)
newStudent.addStudent(face_token) #Add Student to both Face++ and DB
return
def detect_face_dispaly(self):
print('\nDetecting Faces...\n')
face_tokens = detect.detect_face('/home/pi/Desktop/ProfMate/pic.jpg') #use api to check the picture
print('\nFinished\n')
num = len(face_tokens)
print('\nFaces find: ' + str(num) + '\n')
return face_tokens
def load_data(self, img_path):
img_list = []
face = detect_face(img_path)
if face is None:
return None
else:
face_img = st.resize(face, (FACE_CROP_SIZE, FACE_CROP_SIZE))
img_list.append(face_img)
images = np.stack(img_list)
return images
def handle_detect():
"""检查人脸"""
filename = get_filename(offset)
file_path = os.path.join(image_dir, filename)
result = detect_face(face_record.id, file_path)
if result:
detect_label['text'] = u'检测成功'
detect_label['fg'] = 'green'
else:
detect_label['text'] = u'检测失败'
detect_label['fg'] = 'red'
# with np.load("/training/data.npy") as data:
# in_data = data
# with np.load("/training/labels.npy") as labels:
# in_labels = labels
# #Make sure labels and data are same length
# assert in_data.shape[0] == in_labels.shape[0]
# #Create a placeholder for current input instead of copying in all training data
# input_current = tf.placeholder(in_data.dtype, in_data.shape)
# labels_current = tf.placeholder(in_labels.dtype, in_labels.shape)
# dataset = tf.data.Dataset.from_tensor_slices((input_current, labels_current))
# iterator = dataset.make_initializable_iterator()
in_data, labels = detect_face()
#Initialize model based on architecture we desire
NN = Neuron_Network(input_layer, first_hidden, second_hidden, output_layer)
#Compile model by setting all correct parameters
NN.model.compile(optimizer="sgd",
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
NN.model.fit(x=in_data, y=labels, epochs=epoch)
#Call method Fit in order to begin training.
# NN.model.fit(iterator, epochs=epoch,
# steps_per_epoch=in_data.shape[0],
# validations_split=0.2)
