import tensorflow as tf
from FaceAging import FaceAging
from os import environ
import argparse
import model as M
M.set_gpu('1')
environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
def str2bool(v):
if v.lower() in ('yes', 'true', 't', 'y', '1'):
return True
elif v.lower() in ('no', 'false', 'f', 'n', '0'):
return False
else:
raise argparse.ArgumentTypeError('Boolean value expected.')
parser = argparse.ArgumentParser(description='CAAE')
parser.add_argument('--is_train', type=str2bool, default=True)
parser.add_argument('--epoch', type=int, default=250, help='number of epochs')
parser.add_argument('--dataset', type=str, default='UTKFace', help='training dataset name that stored in ./data')
parser.add_argument('--savedir', type=str, default='save', help='dir of saving checkpoints and intermediate training results')
parser.add_argument('--testdir', type=str, default='None', help='dir of testing images')
parser.add_argument('--use_trained_model', type=str2bool, default=True, help='whether train from an existing model or from scratch')
parser.add_argument('--use_init_model', type=str2bool, default=True, help='whether train from the init model if cannot find an existing model')
FLAGS = parser.parse_args()
import tensorflow as tf
from FaceAging import FaceAging
from ops import *
from scipy.misc import imread, imresize, imsave
import model as M
M.set_gpu('0')
import os
if not os.path.exists('./res/'):
os.mkdir('./res/')
sess = tf.Session()
net = FaceAging(sess, is_training=False, size_batch=5)
net.load_checkpoint()
def generate(img, sample_gender, pref):
for i in range(10):
age = np.ones([len(img), 10], dtype=np.float32)
age = age * -1.
age[:, i] = 1.
gen = net.session.run(net.G,
feed_dict={
net.input_image: img,
net.age: age,
net.gender: sample_gender
})
gen = gen + 1.
gen = gen * 0.5
import network as N
import numpy as np
import model as M
import tensorflow as tf
import cv2
M.set_gpu('2')
class recon():
def __init__(self, step):
with tf.variable_scope('Input_holders'):
self.inpholder = tf.placeholder(tf.float32,
[None, step, 256, 256, 3])
self.step = step
self.targetholder = tf.placeholder(tf.float32,
[None, step, 256, 256, 3])
with tf.variable_scope('Head'):
inp_split = tf.unstack(self.inpholder, axis=1)
features = []
for i in range(len(inp_split)):
features.append(N.conv_layers(inp_split[i], i != 0))
features = tf.stack(features, axis=1)
lstm_out = M.SimpleLSTM(4 * 4 * 128).apply(features)
with tf.variable_scope('Tail'):
feat_split = tf.unstack(lstm_out, axis=1)
# I try the last frame for now
feat = feat_split[-1]
A, C = N.deconv_layers(feat)
import AIM_gen
import tensorflow as tf
import numpy as np
import model as M
import cv2
M.set_gpu('3')
aim_mod = AIM_gen.AIM_gen()
# get data_reader
import data_reader
data_reader = data_reader.data_reader('outpt.txt')
BSIZE = 32
ITER_PER_EPOC = 200000 // BSIZE
EPOC = 50
MAX_ITER = ITER_PER_EPOC * EPOC
ETA = M.ETA(MAX_ITER)
ETA.start()
for iteration in range(MAX_ITER + 1):
img, age_fake, age = data_reader.get_train_batch(BSIZE)
losses = aim_mod.train(img, age_fake, age, normalize=True)
if iteration % 10 == 0:
print('------ Iteration %d ---------' % iteration)
aim_mod.display_losses(losses)
print('ETA', ETA.get_ETA(iteration))
if iteration % 1000 == 0 and iteration > 0:
aim_mod.save('%d.ckpt' % iteration)