def build_model(sess, embedding_dim, batch_size):
model = CondGAN(lr_imsize=cfg.TEST.LR_IMSIZE, hr_lr_ratio=int(cfg.TEST.HR_IMSIZE/cfg.TEST.LR_IMSIZE))
embeddings = tf.placeholder(tf.float32, [batch_size, embedding_dim], name='conditional_embeddings')
with tf.variable_scope("g_net"):
c = sample_encoded_context(embeddings, model)
z = tf.random_normal([batch_size, cfg.Z_DIM])
fake_images = model.get_generator(tf.concat([c, z], 1,), False)
with tf.variable_scope("hr_g_net"):
hr_c = sample_encoded_context(embeddings, model)
hr_fake_images = model.hr_get_generator(fake_images, hr_c, False)
ckt_path = cfg.TEST.PRETRAINED_MODEL
if ckt_path.find('.ckpt') != -1:
print("Reading model parameters from %s" % ckt_path)
saver = tf.train.Saver(tf.all_variables())
saver.restore(sess, ckt_path)
else:
print("Input a valid model path.")
return embeddings, fake_images, hr_fake_images
import torch import matplotlib.pyplot as plt from torchvision import utils from model import CondGAN from utils import create_demo, create_c_demo pretraiend = '../pre_trained/CondDCGAN_epoch300.pth.tar' checkpoints = torch.load(pretraiend) net = CondGAN(use_sigmoid=True) net.load_state_dict(checkpoints['state_dict']) z = torch.randn(4, 100, 1, 1) hair, eyes = range(4), range(4) c = torch.from_numpy(create_c_demo(hair, eyes)).float() img = create_demo(net.G, z, c, use_cuda=True, cond=True) img = utils.make_grid(img.data, normalize=True) img = img.cpu().numpy().transpose((1, 2, 0)) plt.imshow(img) plt.show()
timestamp = now.strftime('%Y_%m_%d_%H_%M_%S')
datadir = 'Data/%s' % cfg.DATASET_NAME
dataset = TextDataset(datadir, cfg.EMBEDDING_TYPE, 1)
filename_test = '%s/test' % (datadir)
dataset.test = dataset.get_data(filename_test)
if cfg.TRAIN.FLAG:
filename_train = '%s/train' % (datadir)
dataset.train = dataset.get_data(filename_train)
ckt_logs_dir = "ckt_logs/%s/%s_%s" % (cfg.DATASET_NAME,
cfg.CONFIG_NAME, timestamp)
mkdir_p(ckt_logs_dir)
else:
s_tmp = cfg.TRAIN.PRETRAINED_MODEL
ckt_logs_dir = s_tmp[:s_tmp.find('.ckpt')]
model = CondGAN(image_shape=dataset.image_shape)
algo = CondGANTrainer(model=model,
dataset=dataset,
ckt_logs_dir=ckt_logs_dir)
if cfg.TRAIN.FLAG:
algo.train()
else:
''' For every input text embedding/sentence in the
training and test datasets, generate cfg.TRAIN.NUM_COPY
images with randomness from noise z and conditioning augmentation.'''
algo.evaluate()
'gan': gan,
'a': 0,
'b': 1,
'c': 1,
'clip': 0.01,
'valid': True,
'val_step': 200,
'save_dir': 'results/1_3_2/' + gan,
'save_freq': 50,
'save_grad': False,
'visdom': True,
'visdom_iter': True
}
config = Namespace(**config)
model = CondGAN(name='Cond' + config.gan, use_sigmoid=config.gan == 'DCGAN')
model.apply(weights_init_normal)
if config.gan == 'DCGAN':
optimizer = {
'D': optim.Adam(model.D.parameters(), lr=0.0002, betas=(0.5, 0.999)),
'G': optim.Adam(model.G.parameters(), lr=0.0002, betas=(0.5, 0.999))
}
criterion = bce_loss
elif config.gan == 'LSGAN':
config.a, config.b, config.c = 0, 1, 1
optimizer = {
'D': optim.Adam(model.D.parameters(), lr=0.0001, betas=(0.5, 0.999)),
'G': optim.Adam(model.G.parameters(), lr=0.0001, betas=(0.5, 0.999))
}
criterion = ls_loss
datadir = 'Data/%s' % cfg.DATASET_NAME
dataset = TextDataset(datadir, cfg.EMBEDDING_TYPE, 4)
filename_test = '%s/test' % (datadir)
dataset.test = dataset.get_data(filename_test)
if cfg.TRAIN.FLAG:
filename_train = '%s/train' % (datadir)
dataset.train = dataset.get_data(filename_train)
ckt_logs_dir = "ckt_logs/%s/%s_%s" % \
(cfg.DATASET_NAME, cfg.CONFIG_NAME, timestamp)
mkdir_p(ckt_logs_dir)
else:
s_tmp = cfg.TRAIN.PRETRAINED_MODEL
ckt_logs_dir = s_tmp[:s_tmp.find('.ckpt')]
model = CondGAN(lr_imsize=int(dataset.image_shape[0] /
dataset.hr_lr_ratio),
hr_lr_ratio=dataset.hr_lr_ratio)
algo = CondGANTrainer(model=model,
dataset=dataset,
ckt_logs_dir=ckt_logs_dir)
if cfg.TRAIN.FLAG:
algo.train()
else:
''' For every input text embedding/sentence in the
training and test datasets, generate cfg.TRAIN.NUM_COPY
images with randomness from noise z and conditioning augmentation.'''
algo.evaluate()
from trainer import CondGANTrainer
from misc.get_configs import parse_args
from misc.utils import mkdir_p
if __name__ == "__main__":
args = parse_args()
print(args)
now = datetime.datetime.now()
timestamp = now.strftime('%Y_%m_%d_%H_%M_%S')
dataset = TextDataset(datadir='datasets/' + args.dataset + '/')
print("Dataset created!")
dataset.train = dataset.get_data()
model = CondGAN(args, image_shape=dataset.image_shape)
print("model created!")
# if args.for_training:
ckt_logs_dir = "ckt_logs/%s" % \
("{}_logs".format(args.dataset))
res_dir = "retrieved_res/%s" % \
("{}_res".format(args.dataset))
mkdir_p(ckt_logs_dir)
mkdir_p(res_dir)
with open(ckt_logs_dir + '/args.txt', 'w') as fid:
fid.write(str(args) + '\n')
algo = CondGANTrainer(args,
model=model,
dataset=dataset,