load_size=load_size,
crop_size=crop_size)
b_test_pool = data.ImageData(sess,
b_test_img_paths,
batch_size,
load_size=load_size,
crop_size=crop_size)
a2b_pool = utils.ItemPool()
b2a_pool = utils.ItemPool()
'''summary'''
summary_writer = tf.summary.FileWriter('./summaries/' + dataset + '_spgan',
sess.graph)
'''saver'''
ckpt_dir = './checkpoints/' + dataset + '_spgan'
utils.mkdir(ckpt_dir + '/')
saver = tf.train.Saver(max_to_keep=30)
ckpt_path = utils.load_checkpoint(ckpt_dir, sess, saver)
if ckpt_path is None:
sess.run(tf.global_variables_initializer())
else:
print('Copy variables from % s' % ckpt_path)
'''train'''
try:
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
batch_epoch = min(len(a_data_pool), len(b_data_pool)) // batch_size
max_it = epoch * batch_epoch
now = time.strftime("%c")
"--subclass",
help="predict subclasses",
action="store_true",
)
parser.add_argument(
"--pretrained",
help="Use pre-trained models from the modelzoo",
action="store_true",
)
parser.add_argument(
"--distributed",
help="Use distributed gpu to train models",
action="store_true",
)
# distributed training parameters
parser.add_argument('--world-size',
default=1,
type=int,
help='number of distributed processes')
parser.add_argument('--dist-url',
default='env://',
help='url used to set up distributed training')
args = parser.parse_args()
if args.output_dir:
utils.mkdir(args.output_dir)
main(args)
if False:
pretrainer = PreTrainer(pretrain_config)
for i in range(len(graph.ViewData)):
pretrainer.pretrain(graph.ViewData[i], 'V' + str(i + 1))
model_config = {
'weight_decay': 1.0,
'View_num': View_num,
'View': layers,
'is_init': True,
'pretrain_params_path':
'./Log/' + dataset_name + '/pretrain_params.pkl'
}
pathResult = './result/' + dataset_name + "/"
mkdir(pathResult)
with open(
pathResult + dataset_name + "_" + str(learning_rate) + '_' +
str(beta_W) + '.txt', "w") as f:
# for beta_i in np.transpose([1,10,50,100,200]):
# for alpha_i in [0.001,0.01,0.1,1,10,100]:
# for gama_i in [0.001,0.01,0.1,1,10,100]:
for ccsistent_loss in [0.001, 0.01, 0.1, 0, 1, 10, 50, 100, 200]:
tf.reset_default_graph()
trainer_config = {
'ccsistent_loss':
ccsistent_loss,
'beta_W':
beta_W,
'View_num':
View_num,
ckpt_path = utils.load_checkpoint('./checkpoints/' + dataset + '_spgan',
sess, saver)
saver.restore(sess, ckpt_path)
if ckpt_path is None:
raise Exception('No checkpoint!')
else:
print('Copy variables from % s' % ckpt_path)
#--test--#
b_list = glob('./Datasets/' + dataset + '/bounding_box_train-Market/*.jpg')
a_list = glob('./Datasets/' + dataset + '/bounding_box_train-Duke/*.jpg')
b_save_dir = './test_predictions/' + dataset + '_spgan' + '/bounding_box_train_market2duke/'
a_save_dir = './test_predictions/' + dataset + '_spgan' + '/bounding_box_train_duke2market/'
utils.mkdir([a_save_dir, b_save_dir])
for i in range(len(a_list)):
a_real_ipt = im.imresize(im.imread(a_list[i]), [crop_size, crop_size])
a_real_ipt.shape = 1, crop_size, crop_size, 3
a2b_opt = sess.run(a2b, feed_dict={a_real: a_real_ipt})
a_img_opt = a2b_opt
img_name = os.path.basename(a_list[i])
img_name = 'market_' + img_name # market_style
im.imwrite(im.immerge(a_img_opt, 1, 1), a_save_dir + img_name)
print('Save %s' % (a_save_dir + img_name))
for i in range(len(b_list)):
b_real_ipt = im.imresize(im.imread(b_list[i]), [crop_size, crop_size])
b_real_ipt.shape = 1, crop_size, crop_size, 3
b_data_pool = data.ImageData(sess, b_img_paths, batch_size, load_size=load_size, crop_size=crop_size)
a_test_img_paths = glob('./Datasets/' + dataset + '/bounding_box_train-Duke/*.jpg')
b_test_img_paths = glob('./Datasets/' + dataset + '/bounding_box_train-Market/*.jpg')
a_test_pool = data.ImageData(sess, a_test_img_paths, batch_size, load_size=load_size, crop_size=crop_size)
b_test_pool = data.ImageData(sess, b_test_img_paths, batch_size, load_size=load_size, crop_size=crop_size)
a2b_pool = utils.ItemPool()
b2a_pool = utils.ItemPool()
'''summary'''
summary_writer = tf.summary.FileWriter('./summaries/' + dataset + '_spgan' , sess.graph)
'''saver'''
ckpt_dir = './checkpoints/' + dataset + '_spgan'
utils.mkdir(ckpt_dir + '/')
saver = tf.train.Saver(max_to_keep= 30)
ckpt_path = utils.load_checkpoint(ckpt_dir, sess, saver)
if ckpt_path is None:
sess.run(tf.global_variables_initializer())
else:
print('Copy variables from % s' % ckpt_path)
'''train'''
try:
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
batch_epoch = min(len(a_data_pool), len(b_data_pool)) // batch_size
max_it = epoch * batch_epoch
saver = tf.train.Saver()
ckpt_path = utils.load_checkpoint('./checkpoints/' + dataset + '_spgan', sess, saver)
saver.restore(sess, ckpt_path)
if ckpt_path is None:
raise Exception('No checkpoint!')
else:
print('Copy variables from % s' % ckpt_path)
#--test--#
b_list = glob('./Datasets/' + dataset + '/bounding_box_train-Market/*.jpg')
a_list = glob('./Datasets/' + dataset + '/bounding_box_train-Duke/*.jpg')
b_save_dir = './test_predictions/' + dataset + '_spgan' + '/bounding_box_train_market2duke/'
a_save_dir = './test_predictions/' + dataset + '_spgan' + '/bounding_box_train_duke2market/'
utils.mkdir([a_save_dir, b_save_dir])
for i in range(len(a_list)):
a_real_ipt = im.imresize(im.imread(a_list[i]), [crop_size, crop_size])
a_real_ipt.shape = 1, crop_size, crop_size, 3
a2b_opt = sess.run(a2b, feed_dict={a_real: a_real_ipt})
a_img_opt = a2b_opt
img_name = os.path.basename(a_list[i])
img_name = 'market_'+img_name # market_style
im.imwrite(im.immerge(a_img_opt, 1, 1), a_save_dir + img_name)
print('Save %s' % (a_save_dir + img_name))