def argparser(is_train=True):
def str2bool(v):
return v.lower() == 'true'
parser = argparse.ArgumentParser(
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--debug', action='store_true', default=False)
parser.add_argument('--prefix', type=str, default='default')
parser.add_argument('--train_dir', type=str)
parser.add_argument('--checkpoint', type=str, default=None)
parser.add_argument('--dataset', type=str, default='CIFAR10',
choices=['MNIST', 'SVHN', 'CIFAR10'])
parser.add_argument('--dump_result', type=str2bool, default=False)
# Model
parser.add_argument('--batch_size', type=int, default=32)
parser.add_argument('--n_z', type=int, default=128)
parser.add_argument('--norm_type', type=str, default='batch',
choices=['batch', 'instance', 'None'])
parser.add_argument('--deconv_type', type=str, default='bilinear',
choices=['bilinear', 'nn', 'transpose'])
# Training config {{{
# ========
# log
parser.add_argument('--log_step', type=int, default=10)
parser.add_argument('--write_summary_step', type=int, default=100)
parser.add_argument('--ckpt_save_step', type=int, default=10000)
parser.add_argument('--test_sample_step', type=int, default=100)
parser.add_argument('--output_save_step', type=int, default=1000)
# learning
parser.add_argument('--max_sample', type=int, default=5000,
help='num of samples the model can see')
parser.add_argument('--max_training_steps', type=int, default=10000000)
parser.add_argument('--learning_rate_g', type=float, default=1e-4)
parser.add_argument('--learning_rate_d', type=float, default=1e-4)
parser.add_argument('--update_rate', type=int, default=1)
# }}}
# Testing config {{{
# ========
parser.add_argument('--data_id', nargs='*', default=None)
# }}}
config = parser.parse_args()
dataset_path = os.path.join('./datasets', config.dataset.lower())
dataset_train, dataset_test = dataset.create_default_splits(dataset_path)
img, label = dataset_train.get_data(dataset_train.ids[0])
config.h = img.shape[0]
config.w = img.shape[1]
config.c = img.shape[2]
config.num_class = label.shape[0]
# --- create model ---
model = Model(config, debug_information=config.debug, is_train=is_train)
return config, model, dataset_train, dataset_test
def main():
dataset_path = "./datasets/mnist/"
dataset_train, dataset_test = dataset.create_default_splits(dataset_path)
print(dataset_train)
print(len(dataset_train))
img, label = dataset_train.get_data(dataset_train.ids[0])
h = img.shape[0]
w = img.shape[1]
c = img.shape[2]
num_class = label.shape[0]
print(h, w, c)
def argparser(is_train=True):
def str2bool(v):
return v.lower() == 'true'
parser = argparse.ArgumentParser(
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--debug', action='store_true', default=False)
parser.add_argument('--prefix', type=str, default='default')
parser.add_argument('--train_dir', type=str)
parser.add_argument('--checkpoint', type=str, default=None)
parser.add_argument('--dataset',
type=str,
default=None,
choices=[
"mean_excep_nl_nn", "mean_std_all", "min_max_all",
"raw", "raw_excep_nl_n"
])
parser.add_argument('--dump_result', type=str2bool,
default=False) #保存保存点生成图像的h5py
# Model
parser.add_argument('--batch_size', type=int, default=32)
parser.add_argument('--n_z', type=int, default=128) #噪声维度
parser.add_argument('--norm_type',
type=str,
default='batch',
choices=['batch', 'instance', 'None'])
parser.add_argument('--deconv_type',
type=str,
default='bilinear',
choices=['bilinear', 'nn', 'transpose'])
# Training config {{{
# ========
# log
parser.add_argument('--log_step', type=int, default=10)
parser.add_argument('--write_summary_step', type=int, default=200)
#parser.add_argument('--ckpt_save_step', type=int, default=1)
parser.add_argument('--test_sample_step', type=int,
default=200) #这个训练阶段的测试集就是验证集
parser.add_argument('--output_save_step', type=int, default=500)
# learning
#parser.add_argument('--max_sample', type=int, default=200,
# help='num of samples the model can see')
parser.add_argument('--max_training_steps', type=int, default=30001)
parser.add_argument('--learning_rate_g', type=float, default=2e-4)
parser.add_argument('--learning_rate_d', type=float, default=1e-4)
parser.add_argument(
'--update_rate', type=int,
default=2) #这不是学习率的改变，而是决定生成器和辨别器训练时机的参数（比如训练一次生成器再训练一次辨别器）。
# }}}
# Testing config {{{
# ========
parser.add_argument('--data_id', nargs='*', default=None)
# }}}
config = parser.parse_args()
dataset_path = os.path.join(
'drive/Colab_Notebooks/SSGAN-Tensorflow-master/datasets',
config.dataset.lower())
dataset_train, dataset_test = dataset.create_default_splits(dataset_path)
dataset_train_unlabel, _ = dataset.create_default_splits_unlabel(
dataset_path)
img, label = dataset_train.get_data(dataset_train.ids[0])
config.h = img.shape[0]
config.w = img.shape[1]
config.c = img.shape[2]
config.num_class = label.shape[0] # 5
# --- create model ---
model = Model(config, debug_information=config.debug, is_train=is_train)
return config, model, dataset_train, dataset_train_unlabel, dataset_test
def argparser(is_train=True):
def str2bool(v):
return v.lower() == 'true'
parser = argparse.ArgumentParser(
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--debug', action='store_true', default=False)
parser.add_argument('--prefix', type=str, default='default')
parser.add_argument('--train_dir', type=str)
parser.add_argument('--checkpoint', type=str, default=None)
parser.add_argument('--checkpoint_g', type=str, default=None)
parser.add_argument('--checkpoint_d', type=str, default=None)
parser.add_argument('--dataset', type=str, default='celeba')
parser.add_argument('--dataset_path', type=str, default=None)
parser.add_argument('--img_h', type=int, default=256)
parser.add_argument('--img_w', type=int, default=256)
# Model
parser.add_argument('--batch_size', type=int, default=32)
parser.add_argument('--gan_type', type=str, default='wgan-gp',
choices=['lsgan', 'hinge', 'wgan-gp'])
parser.add_argument('--n_z', type=int, default=128)
parser.add_argument('--num_dis_conv', type=int, default=6)
parser.add_argument('--num_g_res_block', type=int, default=3)
parser.add_argument('--num_d_res_block', type=int, default=3)
parser.add_argument('--g_norm_type', type=str, default='batch',
choices=['batch', 'instance', 'none'])
parser.add_argument('--d_norm_type', type=str, default='none',
choices=['batch', 'instance', 'none'])
parser.add_argument('--deconv_type', type=str, default='bilinear',
choices=['bilinear', 'nn', 'transpose'])
# Training config {{{
# ========
# log
parser.add_argument('--log_step', type=int, default=100)
parser.add_argument('--write_summary_step', type=int, default=100)
parser.add_argument('--ckpt_save_step', type=int, default=10000)
# learning
parser.add_argument('--max_training_steps', type=int, default=10000000)
parser.add_argument('--learning_rate_g', type=float, default=1e-4)
parser.add_argument('--learning_rate_d', type=float, default=1e-4)
parser.add_argument('--adam_beta1', type=float, default=0.5)
parser.add_argument('--adam_beta2', type=float, default=0.9)
parser.add_argument('--lr_weight_decay', type=str2bool, default=False)
parser.add_argument('--update_g', type=int, default=1)
parser.add_argument('--update_d', type=int, default=1)
parser.add_argument('--gamma', type=float, default=10)
# }}}
# Testing config {{{
# ========
parser.add_argument(
'--output_file', type=str, default=None,
help='dump all generated images to a HDF5 file with the filename specify here')
parser.add_argument('--write_summary_image', type=str2bool, default=False)
parser.add_argument('--summary_image_name', type=str, default='summary.png')
parser.add_argument('--max_evaluation_steps', type=int, default=5)
# }}}
config = parser.parse_args()
if config.dataset_path is None:
dataset_path = os.path.join('./datasets', config.dataset.lower())
else:
dataset_path = config.dataset_path
if config.dataset in ['CIFAR10', 'CIFAR100', 'SVHN', 'MNIST', 'Fashion_MNIST']:
import datasets.hdf5_loader as dataset
else:
import datasets.image_loader as dataset
dataset_train, dataset_test = dataset.create_default_splits(
dataset_path, h=config.img_h, w=config.img_w)
img = dataset_train.get_data(dataset_train.ids[0])
config.h = img.shape[0]
config.w = img.shape[1]
config.c = img.shape[2]
# --- create model ---
model = Model(config, debug_information=config.debug, is_train=is_train)
return config, model, dataset_train, dataset_test
def argparser(is_train=True):
def str2bool(v):
return v.lower() == 'true'
parser = argparse.ArgumentParser(
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--debug', action='store_true', default=False)
parser.add_argument('--prefix', type=str, default='default')
parser.add_argument('--train_dir', type=str)
parser.add_argument('--checkpoint', type=str, default=None)
parser.add_argument('--dataset',
type=str,
default='MNIST',
choices=['MNIST', 'SVHN', 'CIFAR10', 'ulna', 'MRI'])
parser.add_argument('--hdf5FileName',
type=str,
default='MRIdata_3_AD_MCI_Normal.hdf5',
choices=[
'MRIdata_2_AD_MCI.hdf5',
'MRIdata_2_AD_Normal.hdf5',
'MRIdata_2_MCI_Normal.hdf5',
'MRIdata_3_AD_MCI_Normal.hdf5'
])
parser.add_argument('--idFileName',
type=str,
default='MRIdata_3_AD_MCI_Normal_id.txt',
choices=[
'MRIdata_2_AD_MCI_id.txt',
'MRIdata_2_AD_Normal_id.txt',
'MRIdata_2_MCI_Normal_id.txt',
'MRIdata_3_AD_MCI_Normal_id.txt'
])
parser.add_argument('--dump_result', type=str2bool, default=False)
# Model
parser.add_argument('--batch_size', type=int, default=3)
parser.add_argument('--n_z', type=int, default=128)
parser.add_argument('--norm_type',
type=str,
default='batch',
choices=['batch', 'instance', 'None'])
parser.add_argument('--deconv_type',
type=str,
default='bilinear',
choices=['bilinear', 'nn', 'transpose'])
# Training config {{{
# ========
# log
parser.add_argument('--log_step', type=int, default=10)
parser.add_argument('--write_summary_step', type=int, default=100)
parser.add_argument('--ckpt_save_step', type=int, default=50)
parser.add_argument('--test_sample_step', type=int, default=100)
parser.add_argument('--output_save_step', type=int, default=100)
# learning
parser.add_argument('--max_sample',
type=int,
default=5000,
help='num of samples the model can see')
parser.add_argument('--max_training_steps', type=int, default=80000)
parser.add_argument('--learning_rate_g', type=float, default=0.0025)
parser.add_argument('--learning_rate_d', type=float, default=0.005)
parser.add_argument('--update_rate', type=int, default=6)
# }}}
# Testing config {{{
# ========
parser.add_argument('--data_id', nargs='*', default=None)
# }}}
config = parser.parse_args()
dataset_path = os.path.join(
r"/media/wenyu/8d268d3e-37df-4af4-ab98-f5660b2e71a7/wenyu/PycharmProjects/SSGAN-original-Tensorflow/datasets",
config.dataset.lower())
dataset_train, dataset_test = dataset.create_default_splits(
dataset_path,
hdf5FileName=config.hdf5FileName,
idFileName=config.idFileName,
cross_validation_number=10)
#dataset_train, dataset_test are 10 cross validation data.
#dataset_train[i] is the i-th fold data
print("step2")
img, label = dataset_train[0].get_data(dataset_train[0].ids[0])
print("step3")
config.h = img.shape[0]
config.w = img.shape[1]
if len(img.shape) == 3:
config.c = img.shape[2]
else:
config.c = 1
config.num_class = label.shape[0]
# --- create model ---
model = Model(config, debug_information=config.debug, is_train=is_train)
return config, model, dataset_train, dataset_test