output_dir = ''
with tf.Graph().as_default():
with tf.Session() as sess:
# embedding_output = tf.placeholder(tf.float32, shape=(None, 128))
# Load model
model = InceptionResnetV1Model()
# Convert to classifier
# model.convert_to_classifier()
# face1 和 face2 可能是同人也可能不是同人，根据lables的值，True/[1,0]：同人；False/[0,1]表示非同人
#目标攻击时加载数据
# faces1, faces2, labels, filepaths1, filepaths2 = set_loader.load_construct_testset_1(200)
faces1, faces2, labels, filepaths1, filepaths2, label_index = set_loader.load_testset(
16)
num_examples = len(faces1)
batch_number = int(math.ceil(num_examples / batch_size))
# thresholds = np.arange(0, 4, 0.01)
threshold = 1.1
# x_adv = []
accuracy_1 = 0
accuracy_2 = 0
accuracy_3 = 0
accuracy_4 = 0
accuracy_5 = 0
# generate the adversarial examples
for ibatch in range(batch_number):
bstart = ibatch * batch_size
bend = min(bstart + batch_size, num_examples)
print('batch size: {}'.format(bend - bstart))
self.layers.append(self.softmax_output)
self.layer_names.append('probs')
def fprop(self, x, set_ref=False):
return dict(zip(self.layer_names, self.layers))
with tf.Graph().as_default():
with tf.Session() as sess:
# Load model
model = InceptionResnetV1Model()
# Convert to classifier
model.convert_to_classifier()
# Load pairs of faces and their labels in one-hot encoding
faces1, faces2, labels = set_loader.load_testset(1000)
# Create victims' embeddings using Facenet itself
graph = tf.get_default_graph()
phase_train_placeholder = graph.get_tensor_by_name("phase_train:0")
feed_dict = {model.face_input: faces2, phase_train_placeholder: False}
victims_embeddings = sess.run(model.embedding_output,
feed_dict=feed_dict)
# Define FGSM for the model
steps = 1
eps = 0.01
alpha = eps / steps
fgsm = FastGradientMethod(model)
fgsm_params = {'eps': alpha, 'clip_min': 0., 'clip_max': 1.}
adv_x = fgsm.generate(model.face_input, **fgsm_params)
self.layer_names.append('logits')
def fprop(self, x, set_ref=False):
return dict(zip(self.layer_names, self.layers))
with tf.Graph().as_default():
with tf.Session() as sess:
# Load model
model = InceptionResnetV1Model()
# Convert to classifier
model.convert_to_classifier()
# Load pairs of faces and their labels in one-hot encoding
size = 100
faces1, faces2, labels = set_loader.load_testset(size)
# Create victims' embeddings using Facenet itself
graph = tf.get_default_graph()
phase_train_placeholder = graph.get_tensor_by_name("phase_train:0")
feed_dict = {model.face_input: faces2, phase_train_placeholder: False}
victims_embeddings = sess.run(model.embedding_output,
feed_dict=feed_dict)
# Define FGSM for the model
steps = 1
eps = 0.01
alpha = eps / steps
fgsm = FastGradientMethod(model)
fgsm_params = {'eps': alpha, 'clip_min': 0., 'clip_max': 1.}
adv_x = fgsm.generate(model.face_input, **fgsm_params)