def __init__(self, sess, args):
self.session = sess
self.image_size = 160
facenet2.load_model(args.model)
# Get input and output tensors
self.images_placeholder = tf.get_default_graph().get_tensor_by_name(
"input:0")
self.embeddings = tf.get_default_graph().get_tensor_by_name(
"embeddings:0")
self.phase_train_placeholder = tf.get_default_graph(
).get_tensor_by_name("phase_train:0")
self.embedding_size = self.embeddings.get_shape()[1]
def __init__(self, proxy_map):
super(SpecificWorker, self).__init__(proxy_map)
### Checking if embedding file exists or not
files = os.listdir('.')
self.file_name = 'data_embeddings.npy'
if (self.file_name in files):
print('Embeddings file found')
self.neural_embeddings = np.load(self.file_name, allow_pickle=True)
else:
### Creating an embedding file if no embeddings exist already
print(
'No File Found. Creating an empty numpy file for neural embeddings'
)
self.neural_embeddings = np.zeros((1, 2), dtype=np.ndarray)
self.neural_embeddings[0, 0] = np.zeros((1, 512), dtype=np.float32)
self.neural_embeddings[0, 1] = '####'
np.save(self.file_name, self.neural_embeddings)
self.timer.timeout.connect(self.compute)
self.Period = 50
self.timer.start(self.Period)
##### Defining the face recognition model and parameters
self.model_path = './assets/20180408-102900/'
##### Thresholds for matching the neural embeddings for label prediction
self.threshold_1 = 1.10
self.threshold_2 = 0.55
##### Relaxed threshold to check if incorrect label is given
self.threshold_3 = 1.25
##### Threshold to remove redundant data
self.threshold_4 = 3
self.image_size = 160
#### Loading the Face Recognition Model
with tf.Graph().as_default():
self.config = tf.ConfigProto()
self.config.gpu_options.per_process_gpu_memory_fraction = 0.3
self.sess = tf.InteractiveSession(config=self.config)
facenet.load_model(self.model_path)
self.images_placeholder = tf.get_default_graph(
).get_tensor_by_name("input:0")
self.images_placeholder = tf.image.resize_images(
self.images_placeholder, (self.image_size, self.image_size))
self.embeddings = tf.get_default_graph().get_tensor_by_name(
"embeddings:0")
self.phase_train_placeholder = tf.get_default_graph(
).get_tensor_by_name("phase_train:0")
# print saver.saver_def.filename_tensor_name
# ckpt = tf.train.get_checkpoint_state('./data/model-20170511-185253.ckpt-80000.data-00000-of-00001')
# saver.restore(sess, ckpt.model_checkpoint_path)
# with sess.as_default():
# load_model('./models/20170511-185253')
# model_checkpoint_path = './models/model-20160506.ckpt-500000'
# saver.restore(sess, model_checkpoint_path)
# saver = tf.train.import_meta_graph('./models/20170511-185253/model-20170511-185253.meta')
# saver.restore(sess, './models/20170511-185253/model-20170511-185253.ckpt-80000')
# print('Facenet embedding restore success')
# load_model('./models/20170511-185253/')
with tf.Graph().as_default():
with tf.Session() as sess:
facenet.load_model('./models/20170512-110547/')
images_placeholder = tf.get_default_graph().get_tensor_by_name("input:0")
embeddings = tf.get_default_graph().get_tensor_by_name("embeddings:0")
phase_train_placeholder = tf.get_default_graph().get_tensor_by_name("phase_train:0")
# images_placeholder = tf.placeholder(tf.float32, shape=(batch_size, image_size, image_size, 3), name='input')
# phase_train_placeholder = tf.placeholder(tf.bool, name='phase_train')
# embeddings = network.inference(images_placeholder, pool_type, use_lrn, 1.0, phase_train=phase_train_placeholder)
train_x = []
train_y = []
sess.run(tf.global_variables_initializer())
for index in range(len(keys)):
for x in data[keys[index]]:
x_color = to_rgb(x)
bounding_boxes, _ = detect_face.detect_face(x_color, minsize, pnet, rnet, onet, threshold, factor)
# nrof_faces = bounding_boxes.shape[0] # number of faces
# restore mtcnn model
print 'Creating networks and loading parameters'
gpu_memory_fraction = 0.8
with tf.Graph().as_default():
gpu_options = tf.GPUOptions(
per_process_gpu_memory_fraction=gpu_memory_fraction)
sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options,
log_device_placement=False))
with sess.as_default():
pnet, rnet, onet = detect_face.create_mtcnn(sess, './data/')
# restore facenet model
print('Restore facenet embedding model')
with tf.Graph().as_default():
with tf.Session() as sess:
facenet.load_model('/home/ubuntu/Code/face/models/20170511-185253/')
images_placeholder = tf.get_default_graph().get_tensor_by_name(
"input:0")
embeddings = tf.get_default_graph().get_tensor_by_name("embeddings:0")
phase_train_placeholder = tf.get_default_graph().get_tensor_by_name(
"phase_train:0")
print('Facenet embedding success')
model = joblib.load(
'/home/ubuntu/Code/face/knn_20170512-110547_2.model')
video_capture = cv2.VideoCapture(2)
# c = 0
find_result = 'other'
keys = update_dir('./train_data/')
sess.run(tf.global_variables_initializer())
def main(args):
print('Creating networks and loading parameters')
with tf.Graph().as_default():
gpu_options = tf.GPUOptions(
per_process_gpu_memory_fraction=args.gpu_memory_fraction)
sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options,
log_device_placement=False))
with sess.as_default():
pnet, rnet, onet = align.detect_face2.create_mtcnn(sess, None)
minsize = 20 # minimum size of face
threshold = [0.6, 0.7, 0.7] # three steps's threshold
factor = 0.709 # scale factor
try:
img = cv2.imread(args.input_image)
except (IOError, ValueError, IndexError) as e:
errorMessage = '{}: {}'.format(args.input_image, e)
print(errorMessage)
return
if img.ndim < 2:
print('Unable to align "%s"' % image_path)
text_file.write('%s\n' % (output_filename))
return
if img.ndim == 2:
img = facenet2.to_rgb(img)
img = img[:, :, 0:3]
bounding_boxes, _ = align.detect_face2.detect_face(
img, minsize, pnet, rnet, onet, threshold, factor)
nrof_faces = bounding_boxes.shape[0]
detected_faces = []
detected_bb = []
if nrof_faces > 0:
det = bounding_boxes[:, 0:4]
det_arr = []
img_size = np.asarray(img.shape)[0:2]
if nrof_faces > 1:
for i in range(nrof_faces):
det_arr.append(np.squeeze(det[i]))
for i, det in enumerate(det_arr):
det = np.squeeze(det)
bb = np.zeros(4, dtype=np.int32)
bb[0] = np.maximum(det[0] - args.margin / 2, 0)
bb[1] = np.maximum(det[1] - args.margin / 2, 0)
bb[2] = np.minimum(det[2] + args.margin / 2, img_size[1])
bb[3] = np.minimum(det[3] + args.margin / 2, img_size[0])
cropped = img[bb[1]:bb[3], bb[0]:bb[2], :]
#scaled = scipy.misc.imresize(cropped, (args.image_size, args.image_size), interp='bilinear')
scaled = cv2.resize(cropped,
(args.image_size, args.image_size))
detected_faces.append(scaled)
detected_bb.append(bb)
print(nrof_faces, 'faces are detected')
print(detected_bb)
source_image = img
print('Loading feature extraction model')
facenet2.load_model(args.model)
# Get input and output tensors
images_placeholder = tf.get_default_graph().get_tensor_by_name(
"input:0")
embeddings = tf.get_default_graph().get_tensor_by_name(
"embeddings:0")
phase_train_placeholder = tf.get_default_graph(
).get_tensor_by_name("phase_train:0")
embedding_size = embeddings.get_shape()[1]
# Run forward pass to calculate embeddings
print('Calculating features for images')
# preprocessing
images = []
for img in detected_faces:
if img.ndim == 2:
img = facenet2.to_rgb(img)
img = facenet2.prewhiten(img)
images.append(img)
feed_dict = {
images_placeholder: images,
phase_train_placeholder: False
}
emb_array = sess.run(embeddings, feed_dict=feed_dict)
print('Testing classifier')
classifier_filename_exp = os.path.expanduser(
args.classifier_filename)
with open(classifier_filename_exp, 'rb') as infile:
(model, class_names) = pickle.load(infile)
print('Loaded classifier model from file "%s"' %
classifier_filename_exp)
predictions = model.predict_proba(emb_array)
best_class_indices = np.argmax(predictions, axis=1)
best_class_probabilities = predictions[
np.arange(len(best_class_indices)), best_class_indices]
for i in range(len(best_class_indices)):
print('%4d %s: %.3f' % (i, class_names[best_class_indices[i]],
best_class_probabilities[i]))
source_image = np.array(source_image)
source_image = cv2.cvtColor(source_image, cv2.COLOR_BGR2RGB)
for i in range(len(detected_bb)):
bb = detected_bb[i]
cv2.rectangle(source_image, (bb[0], bb[1]), (bb[2], bb[3]),
(0, 255, 0), 3)
cv2.putText(source_image, class_names[best_class_indices[i]],
(bb[0], bb[1] - 20), cv2.FONT_HERSHEY_SIMPLEX, 1,
(0, 255, 0), 2)
cv2.putText(source_image, '%.3f' % best_class_probabilities[i],
(bb[0], bb[3] + 40), cv2.FONT_HERSHEY_SIMPLEX, 1,
(0, 255, 0), 2)
cv2.imwrite('result.png', source_image)