Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

To support more datasets #53

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

To support more datasets #53

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f16f009
To support more datasets
ai-porter Dec 11, 2018
73bf012
To support more datasets
ai-porter Dec 11, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 6 additions & 4 deletions main.py
100644 → 100755
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,8 +10,10 @@
parser.add_argument('--epoch_step', dest='epoch_step', type=int, default=100, help='# of epoch to decay lr')
parser.add_argument('--batch_size', dest='batch_size', type=int, default=1, help='# images in batch')
parser.add_argument('--train_size', dest='train_size', type=int, default=1e8, help='# images used to train')
parser.add_argument('--load_size', dest='load_size', type=int, default=286, help='scale images to this size')
parser.add_argument('--fine_size', dest='fine_size', type=int, default=256, help='then crop to this size')
parser.add_argument('--load_height', dest='load_height', type=int, default=258, help='scale images to this size')
parser.add_argument('--load_width', dest='load_width', type=int, default=514, help='scale images to this size')
parser.add_argument('--fine_height', dest='fine_height', type=int, default=256, help='then crop to this size')
parser.add_argument('--fine_width', dest='fine_width', type=int, default=512, help='then crop to this size')
parser.add_argument('--ngf', dest='ngf', type=int, default=64, help='# of gen filters in first conv layer')
parser.add_argument('--ndf', dest='ndf', type=int, default=64, help='# of discri filters in first conv layer')
parser.add_argument('--input_nc', dest='input_nc', type=int, default=3, help='# of input image channels')
Expand Down Expand Up @@ -42,8 +44,8 @@ def main(_):
if not os.path.exists(args.test_dir):
os.makedirs(args.test_dir)

tfconfig = tf.ConfigProto(allow_soft_placement=True)
tfconfig.gpu_options.allow_growth = True
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction = 0.95, allow_growth = False)
tfconfig = tf.ConfigProto(allow_soft_placement=True, gpu_options = gpu_options)
with tf.Session(config=tfconfig) as sess:
model = cyclegan(sess, args)
model.train(args) if args.phase == 'train' \
Expand Down
55 changes: 55 additions & 0 deletions main_horse.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,55 @@
import argparse
import os
import tensorflow as tf
tf.set_random_seed(19)
from model import cyclegan

parser = argparse.ArgumentParser(description='')
parser.add_argument('--dataset_dir', dest='dataset_dir', default='horse2zebra', help='path of the dataset')
parser.add_argument('--epoch', dest='epoch', type=int, default=200, help='# of epoch')
parser.add_argument('--epoch_step', dest='epoch_step', type=int, default=100, help='# of epoch to decay lr')
parser.add_argument('--batch_size', dest='batch_size', type=int, default=1, help='# images in batch')
parser.add_argument('--train_size', dest='train_size', type=int, default=1e8, help='# images used to train')
parser.add_argument('--load_height', dest='load_height', type=int, default=286, help='scale images to this size')
parser.add_argument('--load_width', dest='load_width', type=int, default=286, help='scale images to this size')
parser.add_argument('--fine_height', dest='fine_height', type=int, default=256, help='then crop to this size')
parser.add_argument('--fine_width', dest='fine_width', type=int, default=256, help='then crop to this size')
parser.add_argument('--ngf', dest='ngf', type=int, default=64, help='# of gen filters in first conv layer')
parser.add_argument('--ndf', dest='ndf', type=int, default=64, help='# of discri filters in first conv layer')
parser.add_argument('--input_nc', dest='input_nc', type=int, default=3, help='# of input image channels')
parser.add_argument('--output_nc', dest='output_nc', type=int, default=3, help='# of output image channels')
parser.add_argument('--lr', dest='lr', type=float, default=0.0002, help='initial learning rate for adam')
parser.add_argument('--beta1', dest='beta1', type=float, default=0.5, help='momentum term of adam')
parser.add_argument('--which_direction', dest='which_direction', default='AtoB', help='AtoB or BtoA')
parser.add_argument('--phase', dest='phase', default='train', help='train, test')
parser.add_argument('--save_freq', dest='save_freq', type=int, default=1000, help='save a model every save_freq iterations')
parser.add_argument('--print_freq', dest='print_freq', type=int, default=100, help='print the debug information every print_freq iterations') # default 100
parser.add_argument('--continue_train', dest='continue_train', type=bool, default=False, help='if continue training, load the latest model: 1: true, 0: false')
parser.add_argument('--checkpoint_dir', dest='checkpoint_dir', default='./checkpoint', help='models are saved here')
parser.add_argument('--sample_dir', dest='sample_dir', default='./sample', help='sample are saved here')
parser.add_argument('--test_dir', dest='test_dir', default='./test', help='test sample are saved here')
parser.add_argument('--L1_lambda', dest='L1_lambda', type=float, default=10.0, help='weight on L1 term in objective')
parser.add_argument('--use_resnet', dest='use_resnet', type=bool, default=True, help='generation network using reidule block')
parser.add_argument('--use_lsgan', dest='use_lsgan', type=bool, default=True, help='gan loss defined in lsgan')
parser.add_argument('--max_size', dest='max_size', type=int, default=50, help='max size of image pool, 0 means do not use image pool')

args = parser.parse_args()


def main(_):
if not os.path.exists(args.checkpoint_dir):
os.makedirs(args.checkpoint_dir)
if not os.path.exists(args.sample_dir):
os.makedirs(args.sample_dir)
if not os.path.exists(args.test_dir):
os.makedirs(args.test_dir)

gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction = 0.95, allow_growth = False)
tfconfig = tf.ConfigProto(allow_soft_placement=True, gpu_options = gpu_options)
with tf.Session(config=tfconfig) as sess:
model = cyclegan(sess, args)
model.train(args) if args.phase == 'train' \
else model.test(args)

if __name__ == '__main__':
tf.app.run()
67 changes: 37 additions & 30 deletions model.py
100644 → 100755
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,8 @@ class cyclegan(object):
def __init__(self, sess, args):
self.sess = sess
self.batch_size = args.batch_size
self.image_size = args.fine_size
self.image_height = args.fine_height
self.image_width = args.fine_width
self.input_c_dim = args.input_nc
self.output_c_dim = args.output_nc
self.L1_lambda = args.L1_lambda
Expand All @@ -30,9 +31,9 @@ def __init__(self, sess, args):
else:
self.criterionGAN = sce_criterion

OPTIONS = namedtuple('OPTIONS', 'batch_size image_size \
OPTIONS = namedtuple('OPTIONS', 'batch_size image_height image_width \
gf_dim df_dim output_c_dim is_training')
self.options = OPTIONS._make((args.batch_size, args.fine_size,
self.options = OPTIONS._make((args.batch_size, args.fine_height, args.fine_width,
args.ngf, args.ndf, args.output_nc,
args.phase == 'train'))

Expand All @@ -42,7 +43,7 @@ def __init__(self, sess, args):

def _build_model(self):
self.real_data = tf.placeholder(tf.float32,
[None, self.image_size, self.image_size,
[None, self.image_height, self.image_width,
self.input_c_dim + self.output_c_dim],
name='real_A_and_B_images')

Expand All @@ -68,10 +69,10 @@ def _build_model(self):
+ self.L1_lambda * abs_criterion(self.real_B, self.fake_B_)

self.fake_A_sample = tf.placeholder(tf.float32,
[None, self.image_size, self.image_size,
[None, self.image_height, self.image_width,
self.input_c_dim], name='fake_A_sample')
self.fake_B_sample = tf.placeholder(tf.float32,
[None, self.image_size, self.image_size,
[None, self.image_height, self.image_width,
self.output_c_dim], name='fake_B_sample')
self.DB_real = self.discriminator(self.real_B, self.options, reuse=True, name="discriminatorB")
self.DA_real = self.discriminator(self.real_A, self.options, reuse=True, name="discriminatorA")
Expand Down Expand Up @@ -104,18 +105,18 @@ def _build_model(self):
)

self.test_A = tf.placeholder(tf.float32,
[None, self.image_size, self.image_size,
[None, self.image_height, self.image_width,
self.input_c_dim], name='test_A')
self.test_B = tf.placeholder(tf.float32,
[None, self.image_size, self.image_size,
[None, self.image_height, self.image_width,
self.output_c_dim], name='test_B')
self.testB = self.generator(self.test_A, self.options, True, name="generatorA2B")
self.testA = self.generator(self.test_B, self.options, True, name="generatorB2A")

t_vars = tf.trainable_variables()
self.d_vars = [var for var in t_vars if 'discriminator' in var.name]
self.g_vars = [var for var in t_vars if 'generator' in var.name]
for var in t_vars: print(var.name)


def train(self, args):
"""Train cyclegan"""
Expand All @@ -127,7 +128,11 @@ def train(self, args):

init_op = tf.global_variables_initializer()
self.sess.run(init_op)
self.writer = tf.summary.FileWriter("./logs", self.sess.graph)

log_dir = "./logs/" + self.dataset_dir
if not os.path.exists(log_dir):
os.makedirs(log_dir)
self.writer = tf.summary.FileWriter(log_dir, self.sess.graph)

counter = 1
start_time = time.time()
Expand All @@ -139,6 +144,7 @@ def train(self, args):
print(" [!] Load failed...")

for epoch in range(args.epoch):
print('The %d epoch starts...'%(epoch+1))
dataA = glob('./datasets/{}/*.*'.format(self.dataset_dir + '/trainA'))
dataB = glob('./datasets/{}/*.*'.format(self.dataset_dir + '/trainB'))
np.random.shuffle(dataA)
Expand All @@ -149,7 +155,7 @@ def train(self, args):
for idx in range(0, batch_idxs):
batch_files = list(zip(dataA[idx * self.batch_size:(idx + 1) * self.batch_size],
dataB[idx * self.batch_size:(idx + 1) * self.batch_size]))
batch_images = [load_train_data(batch_file, args.load_size, args.fine_size) for batch_file in batch_files]
batch_images = [load_train_data(batch_file, args.load_height, args.load_width, args.fine_height, args.fine_width) for batch_file in batch_files]
batch_images = np.array(batch_images).astype(np.float32)

# Update G network and record fake outputs
Expand All @@ -173,14 +179,14 @@ def train(self, args):
epoch, idx, batch_idxs, time.time() - start_time)))

if np.mod(counter, args.print_freq) == 1:
self.sample_model(args.sample_dir, epoch, idx)
self.sample_model(args, args.sample_dir, epoch, idx)

if np.mod(counter, args.save_freq) == 2:
self.save(args.checkpoint_dir, counter)

def save(self, checkpoint_dir, step):
model_name = "cyclegan.model"
model_dir = "%s_%s" % (self.dataset_dir, self.image_size)
model_dir = "%s_%s_%s" % (self.dataset_dir, self.image_height, self.image_width)
checkpoint_dir = os.path.join(checkpoint_dir, model_dir)

if not os.path.exists(checkpoint_dir):
Expand All @@ -193,7 +199,7 @@ def save(self, checkpoint_dir, step):
def load(self, checkpoint_dir):
print(" [*] Reading checkpoint...")

model_dir = "%s_%s" % (self.dataset_dir, self.image_size)
model_dir = "%s_%s_%s" % (self.dataset_dir, self.image_height, self.image_width)
checkpoint_dir = os.path.join(checkpoint_dir, model_dir)

ckpt = tf.train.get_checkpoint_state(checkpoint_dir)
Expand All @@ -204,23 +210,24 @@ def load(self, checkpoint_dir):
else:
return False

def sample_model(self, sample_dir, epoch, idx):
dataA = glob('./datasets/{}/*.*'.format(self.dataset_dir + '/testA'))
dataB = glob('./datasets/{}/*.*'.format(self.dataset_dir + '/testB'))
def sample_model(self, args, sample_dir, epoch, idx):
sample_dir_new = sample_dir + '/' + self.dataset_dir
if not os.path.exists(sample_dir_new):
os.makedirs(sample_dir_new)

dataA = glob('./datasets/{}/*.*'.format(self.dataset_dir + '/trainA'))
dataB = glob('./datasets/{}/*.*'.format(self.dataset_dir + '/trainB'))
np.random.shuffle(dataA)
np.random.shuffle(dataB)
batch_files = list(zip(dataA[:self.batch_size], dataB[:self.batch_size]))
sample_images = [load_train_data(batch_file, is_testing=True) for batch_file in batch_files]
sample_images = np.array(sample_images).astype(np.float32)

fake_A, fake_B = self.sess.run(
[self.fake_A, self.fake_B],
feed_dict={self.real_data: sample_images}
)
save_images(fake_A, [self.batch_size, 1],
'./{}/A_{:02d}_{:04d}.jpg'.format(sample_dir, epoch, idx))
save_images(fake_B, [self.batch_size, 1],
'./{}/B_{:02d}_{:04d}.jpg'.format(sample_dir, epoch, idx))
for ind in range(5): # display 5 result images in each epoch
batch_files = list(zip(dataA[ind * self.batch_size:(ind + 1) * self.batch_size], dataB[ind * self.batch_size:(ind + 1) * self.batch_size]))
sample_images = [load_train_data(batch_file, args.load_height, args.load_width, args.fine_height, args.fine_width, is_testing=True) for batch_file in batch_files]
sample_images = np.array(sample_images).astype(np.float32)
fake_A, fake_B = self.sess.run([self.fake_A, self.fake_B], feed_dict={self.real_data: sample_images})
real_A = sample_images[:, :, :, :self.input_c_dim]
real_B = sample_images[:, :, :, self.input_c_dim:self.input_c_dim + self.output_c_dim]
concat_img = np.concatenate([real_A, real_B, fake_B, fake_A], 2)
save_images(concat_img, [self.batch_size, 1], '{}/output_{:03d}_{:02d}.jpg'.format(sample_dir_new, epoch, ind))

def test(self, args):
"""Test cyclegan"""
Expand Down Expand Up @@ -249,7 +256,7 @@ def test(self, args):

for sample_file in sample_files:
print('Processing image: ' + sample_file)
sample_image = [load_test_data(sample_file, args.fine_size)]
sample_image = [load_test_data(sample_file, args.fine_height, args.fine_width)]
sample_image = np.array(sample_image).astype(np.float32)
image_path = os.path.join(args.test_dir,
'{0}_{1}'.format(args.which_direction, os.path.basename(sample_file)))
Expand Down
1 change: 1 addition & 0 deletions train_horse.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
CUDA_VISIBLE_DEVICES=0 python3 main_horse.py --dataset_dir=horse2zebra
25 changes: 13 additions & 12 deletions utils.py
100644 → 100755
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,35 +42,36 @@ def __call__(self, image):
else:
return image

def load_test_data(image_path, fine_size=256):
def load_test_data(image_path, fine_height=256, fine_width=256):
img = imread(image_path)
img = scipy.misc.imresize(img, [fine_size, fine_size])
img = scipy.misc.imresize(img, [fine_height, fine_width])
img = img/127.5 - 1
return img

def load_train_data(image_path, load_size=286, fine_size=256, is_testing=False):
def load_train_data(image_path, load_height = 286, load_width = 286, fine_height = 256, fine_width = 256, is_testing=False):
img_A = imread(image_path[0])
img_B = imread(image_path[1])
if not is_testing:
img_A = scipy.misc.imresize(img_A, [load_size, load_size])
img_B = scipy.misc.imresize(img_B, [load_size, load_size])
h1 = int(np.ceil(np.random.uniform(1e-2, load_size-fine_size)))
w1 = int(np.ceil(np.random.uniform(1e-2, load_size-fine_size)))
img_A = img_A[h1:h1+fine_size, w1:w1+fine_size]
img_B = img_B[h1:h1+fine_size, w1:w1+fine_size]
img_A = scipy.misc.imresize(img_A, [load_height, load_width])
img_B = scipy.misc.imresize(img_B, [load_height, load_width])
h1 = int(np.ceil(np.random.uniform(1e-2, load_height-fine_height)))
w1 = int(np.ceil(np.random.uniform(1e-2, load_width-fine_width)))
img_A = img_A[h1:h1+fine_height, w1:w1+fine_width]
img_B = img_B[h1:h1+fine_height, w1:w1+fine_width]

if np.random.random() > 0.5:
img_A = np.fliplr(img_A)
img_B = np.fliplr(img_B)
else:
img_A = scipy.misc.imresize(img_A, [fine_size, fine_size])
img_B = scipy.misc.imresize(img_B, [fine_size, fine_size])
img_A = scipy.misc.imresize(img_A, [fine_height, fine_width])
img_B = scipy.misc.imresize(img_B, [fine_height, fine_width])

img_A = img_A/127.5 - 1.
img_B = img_B/127.5 - 1.

img_AB = np.concatenate((img_A, img_B), axis=2)
# img_AB shape: (fine_size, fine_size, input_c_dim + output_c_dim)

# img_AB shape: (fine_height, fine_width, input_c_dim + output_c_dim)
return img_AB

# -----------------------------
Expand Down