def generate_output_file(gtype, outfile):
gd = GenData()
for k in gtype:
output = None
if k == 'c':
template_file = get_file_data("option.c.in")
output = template_file % gd.codec
elif k == 'h':
template_file = get_file_data("option.h.in")
output = template_file % gd.codeh
elif k == 'e':
output = gd.codef
else:
assert (False)
if output is not None:
save_result(outfile, output)
def generate_data(self, generate_size, sample_size, budget = 100):
gen_data_list = []
classfilter = self.__classfilter
# This step can be parallelized for each category
for NUM in classfilter.getSeenClass():
original_data = classfilter.getDatabyLabel()[str(NUM)]
gen_data = GenData(original_data[:sample_size,],class_num = NUM, generate_size = generate_size,classifier = self.__classifier, budget = budget)
#print "Generate positive data of class ", NUM
gen_data.generate_negative_data(dim_range = [0,1])
#print "Generate negative data of class ", NUM
gen_data.generate_positive_data(dim_range = [0,1])
gen_data_list.append(gen_data)
self.__gen_data_list = gen_data_list
self.__plus_label = []
for idx in range(len(self.__gen_data_list)):
self.__plus_label.append(gen_data_list[idx].getClassNum())
import tensorflow as tf
LOGS_Path = "./logs/"
CHECKPOINTS_PATH = './checkpoints2/'
BATCH_SIZE = 8
LEARNING_RATE = .0001
BETA = .75
EXP_NAME = f"beta_{BETA}"
if __name__ == "__main__":
model = Model()
data = GenData('./optimization-ii-project-3/')
files_list = data.files_list
sess = tf.InteractiveSession(graph=tf.Graph(), config=tf.ConfigProto(log_device_placement=True))
secret_tensor = tf.placeholder(shape=[None,224,224,3],dtype=tf.float32,name="input_prep")
cover_tensor = tf.placeholder(shape=[None,224,224,3],dtype=tf.float32,name="input_hide")
global_step_tensor = tf.Variable(0, trainable=False, name='global_step')
train_op , summary_op, loss_op,secret_loss_op,cover_loss_op = model.prepare_training_graph(secret_tensor,cover_tensor,global_step_tensor)
writer = tf.summary.FileWriter(os.path.join(LOGS_Path,EXP_NAME),sess.graph)
test_op, test_loss_op,test_secret_loss_op,test_cover_loss_op = model.prepare_test_graph(secret_tensor,cover_tensor)
covered_tensor = tf.placeholder(shape=[None,224,224,3],dtype=tf.float32,name="deploy_covered")
deploy_hide_image_op , deploy_reveal_image_op = model.prepare_deployment_graph(secret_tensor,cover_tensor,covered_tensor)
import sys
sys.path.append("..")
from common import config
config.GPU = True
from gen_data import GenData
from common.optimizer import Adam
from common.trainer import Trainer
from eval_tools import eval_seq2seq
#from seq2seq import Seq2seq
#from seq2seq.peeky_seq2seq import PeekySeq2seq
from attention.attention_bi_seq2seq import AttentionBiSeq2seq
from common.gpu import to_cpu, to_gpu
d = GenData()
#d.unzip_and_gen_data()
(x_train, t_train), (x_test, t_test) = d.load_corpus(file_name="realdata.txt")
word_to_id_q, word_to_id_a, id_to_word_q, id_to_word_a = d.get_vocab()
is_reverse = True
if is_reverse:
x_train, x_test = x_train[:, ::-1], x_test[:, ::-1]
# ハイパーパラメータ設定
vocab_size_x = len(word_to_id_q)
vocab_size_t = len(word_to_id_a)
wordvec_size = 1000
hidden_size = 1000
batch_size = 50
max_epoch = 20
max_grad = 5.0
# -*- coding: utf-8 -*-
from model import Model
from gen_data import GenData
import os
import tensorflow as tf
import numpy as np
model = Model()
data = GenData('./optimization-ii-project-3/')
files_list = data.files_list
secret_tensor = tf.placeholder(shape=[None,224,224,3],dtype=tf.float32,name="input_prep")
cover_tensor = tf.placeholder(shape=[None,224,224,3],dtype=tf.float32,name="input_hide")
global_step_tensor = tf.Variable(0, trainable=False, name='global_step')
train_op , summary_op, loss_op,secret_loss_op,cover_loss_op = model.prepare_training_graph(secret_tensor,cover_tensor,global_step_tensor)
test_op, test_loss_op,test_secret_loss_op,test_cover_loss_op = model.prepare_test_graph(secret_tensor,cover_tensor)
covered_tensor = tf.placeholder(shape=[None,224,224,3],dtype=tf.float32,name="deploy_covered")
deploy_hide_image_op = model.encode(secret_tensor,cover_tensor)
deploy_reveal_image_op = model.decode(covered_tensor)
saver = tf.train.Saver()
sess = tf.InteractiveSession(config=tf.ConfigProto(log_device_placement=True))
saver.restore(sess, './checkpoints/beta_0.750.1396-2101')
from matplotlib import pyplot as plt