def create_dataset(dataset_opt):
mode = dataset_opt["mode"]
if mode == "LQ": # Predictor
from data.LQ_dataset import LQDataset as D
dataset = D(dataset_opt)
elif mode == "LQGTker": # SFTMD
from data.LQGTker_dataset import LQGTKerDataset as D
dataset = D(dataset_opt)
elif mode == "SRker": # Corrector
from data.SRker_dataset import SRkerDataset as D
dataset = D(dataset_opt)
# elif mode == 'LQGTseg_bg':
# from data.LQGT_seg_bg_dataset import LQGTSeg_BG_Dataset as D
else:
raise NotImplementedError("Dataset [{:s}] is not recognized.".format(mode))
logger = logging.getLogger("base")
logger.info(
"Dataset [{:s} - {:s}] is created.".format(
dataset.__class__.__name__, dataset_opt["name"]
)
)
return dataset
def create_dataset(dataset_opt):
mode = dataset_opt['mode']
# datasets for image restoration
if mode == 'LQ':
from data.LQ_dataset import LQDataset as D
elif mode == 'LQGT':
from data.LQGT_dataset import LQGTDataset as D
elif mode == 'Color':
from data.Color_dataset import ColorDataset as D
elif mode == 'ContinueLQGT':
from data.ContinueLQGT_dataset import ContinueLQGTDataset as D
# datasets for video restoration
elif mode == 'REDS':
from data.REDS_dataset import REDSDataset as D
elif mode == 'Vimeo90K':
from data.Vimeo90K_dataset import Vimeo90KDataset as D
elif mode == 'video_test':
from data.video_test_dataset import VideoTestDataset as D
else:
raise NotImplementedError(
'Dataset [{:s}] is not recognized.'.format(mode))
dataset = D(dataset_opt)
logger = logging.getLogger('base')
logger.info('Dataset [{:s} - {:s}] is created.'.format(
dataset.__class__.__name__, dataset_opt['name']))
return dataset
def create_dataset(dataset_opt):
# assign dataset
# Vimeo90K: Vimeo90K train&val
# video_test: Vid4 test
mode = dataset_opt['mode']
if mode == 'LQ':
from data.LQ_dataset import LQDataset as D
elif mode == 'LQGT':
from data.LQGT_dataset import LQGTDataset as D
elif mode == 'Vimeo90K':
from data.Vimeo90K_dataset import Vimeo90KDataset as D
elif mode == 'video_test':
from data.video_test_dataset import VideoTestDataset as D
elif mode in ['DIV2K_easy', 'DIV2K_train']:
from data.DIV2K_dataset import ImageTrainDataset as D
elif mode in ['DIV2K_val']:
from data.DIV2K_dataset import ImageValDataset as D
else:
raise NotImplementedError(
'Dataset [{:s}] is not recognized.'.format(mode))
dataset = D(dataset_opt)
logger = logging.getLogger('base')
logger.info('Dataset [{:s} - {:s}] is created.'.format(
dataset.__class__.__name__, dataset_opt['name']))
return dataset
def create_dataset(dataset_opt):
mode = dataset_opt['mode']
if mode == 'LQ':
from data.LQ_dataset import LQDataset as D
elif mode == 'LQGT':
from data.LQGT_dataset import LQGTDataset as D
elif mode == 'VISR':
from data.VISR_dataset import VISRDataset as D
elif mode == 'SEV':
from data.SEV_dataset import SEVDataset as D
elif mode == 'REDS':
from data.REDS_dataset import REDSDataset as D
elif mode == 'video_test':
from data.video_test_dataset import VideoTestDataset as D
# elif mode == 'LQGTseg_bg':
# from data.LQGT_seg_bg_dataset import LQGTSeg_BG_Dataset as D
else:
raise NotImplementedError(
'Dataset [{:s}] is not recognized.'.format(mode))
dataset = D(dataset_opt)
logger = logging.getLogger('base')
logger.info('Dataset [{:s} - {:s}] is created.'.format(
dataset.__class__.__name__, dataset_opt['name']))
return dataset
def create_dataset(dataset_opt):
mode = dataset_opt['mode'] # mode ~ which dataset to use
if mode == 'LQ':
from data.LQ_dataset import LQDataset as D
dataset = D(dataset_opt)
elif mode == 'LQGT':
from data.LQGT_dataset import LQGTDataset as D
dataset = D(dataset_opt)
elif mode == 'FastMRI':
from data.fastmri_dataset import FASTMRIDataset as D
from data.fastmri import subsample, transforms
# Create a mask function
mask_func = subsample.RandomMaskFunc(center_fractions=[0.08],
accelerations=[4])
class DataTransform:
def __call__(self, target, mask_func, seed=None):
# Preprocess the data here
# target shape: [H, W, 1] or [H, W, 3]
if target.shape[2] == 1:
img = np.concatenate((target, np.zeros_like(target)),
axis=2)
assert img.shape[-1] == 2
img = transforms.to_tensor(img)
kspace = transforms.fft2(img)
center_kspace, _ = transforms.apply_mask(kspace,
mask_func,
seed=seed)
img_LF = transforms.complex_abs(
transforms.ifft2(center_kspace))
img_LF = img_LF.unsqueeze(0)
# img_LF tensor should have shape [H, W, ?]
target = transforms.to_tensor(np.transpose(
target, (2, 0, 1))) # target shape [1, H, W]
return img_LF, target
dataset = D(dataset_opt, mask_func, transform=DataTransform())
else:
raise NotImplementedError(
'Dataset [{:s}] is not recognized.'.format(mode))
logger = logging.getLogger('base')
logger.info('Dataset [{:s} - {:s}] is created.'.format(
dataset.__class__.__name__, dataset_opt['name']))
return dataset
def create_dataset(dataset_opt, is_train=True):
mode = dataset_opt['mode']
# datasets for image restoration
if mode == 'LQ':
from data.LQ_dataset import LQDataset as D
elif mode == 'LQGT':
from data.LQGT_dataset import LQGTDataset as D
elif mode == 'RANK_IMIM_Pair':
from data.Rank_IMIM_Pair_dataset import RANK_IMIM_Pair_Dataset as D
else:
raise NotImplementedError(
'Dataset [{:s}] is not recognized.'.format(mode))
if 'RANK_IMIM_Pair' in mode:
dataset = D(dataset_opt, is_train=is_train)
else:
dataset = D(dataset_opt)
logger = logging.getLogger('base')
logger.info('Dataset [{:s} - {:s}] is created.'.format(
dataset.__class__.__name__, dataset_opt['name']))
return dataset
def create_dataset(dataset_opt):
mode = dataset_opt['mode']
# datasets for image restoration
if mode == 'LQ':
from data.LQ_dataset import LQDataset as D
elif mode == 'LQGT':
from data.LQGT_dataset import LQGTDataset as D
else:
raise NotImplementedError(
'Dataset [{:s}] is not recognized.'.format(mode))
dataset = D(dataset_opt)
logger = logging.getLogger('base')
logger.info('Dataset [{:s} - {:s}] is created.'.format(
dataset.__class__.__name__, dataset_opt['name']))
return dataset
def create_dataset(dataset_opt):
mode = dataset_opt['mode']
if mode == 'LQ':
from data.LQ_dataset import LQDataset as D
elif mode == 'LQGT':
from data.LQGT_dataset import LQGTDataset as D
elif mode == 'LQGT_nopatch':
from data.LQGT_nopatch_dataset import LQGTDataset as D
# elif mode == 'LQGTseg_bg':
# from data.LQGT_seg_bg_dataset import LQGTSeg_BG_Dataset as D
else:
raise NotImplementedError(
'Dataset [{:s}] is not recognized.'.format(mode))
dataset = D(dataset_opt)
logger = logging.getLogger('base')
logger.info('Dataset [{:s} - {:s}] is created.'.format(
dataset.__class__.__name__, dataset_opt['name']))
return dataset
def create_dataset(dataset_opt):
mode = dataset_opt['mode']
# datasets for image restoration
if mode == 'LQ':
from data.LQ_dataset import LQDataset as D
elif mode == 'LQGT':
from data.LQGT_dataset import LQGTDataset as D
# datasets for video restoration
elif mode == 'REDS':
from data.REDS_dataset import REDSDataset as D
elif mode == 'REDSImg':
from data.REDS_dataset import REDSImgDataset as D
elif mode == 'REDSMultiImg':
from data.REDS_dataset import REDSMultiImgDataset as D
elif mode == 'MultiREDS':
from data.REDS_dataset import MultiREDSDataset as D
elif mode == 'MetaREDS':
from data.REDS_dataset import MetaREDSDataset as D
elif mode == 'MetaREDSOnline':
from data.REDS_dataset import MetaREDSDatasetOnline as D
elif mode == 'UPREDS':
from data.REDS_dataset import UPREDSDataset as D
elif mode == 'Vimeo90K':
from data.Vimeo90K_dataset import Vimeo90KDataset as D
elif mode == 'UPVimeo':
from data.Vimeo90K_dataset import UPVimeoDataset as D
elif mode == 'video_test':
from data.video_test_dataset import VideoTestDataset as D
elif mode == 'online_video_test':
from data.video_test_dataset import OnlineVideoTestDataset as D
elif mode == 'img_test':
from data.video_test_dataset import ImgTestDataset as D
else:
raise NotImplementedError('Dataset [{:s}] is not recognized.'.format(mode))
dataset = D(dataset_opt)
logger = logging.getLogger('base')
logger.info('Dataset [{:s} - {:s}] is created.'.format(dataset.__class__.__name__,
dataset_opt['name']))
return dataset