def get_dataset(all_cfg):
cfg = all_cfg.data
input_transform = transforms.Compose([
sep_transforms.ArrayToTensor(),
transforms.Normalize(mean=[0, 0, 0], std=[255, 255, 255]),
])
co_transform = get_co_transforms(aug_args=all_cfg.data_aug)
if cfg.type == 'KITTI_15':
train_input_transform = copy.deepcopy(input_transform)
train_input_transform.transforms.insert(
0, sep_transforms.Zoom(*cfg.train_shape))
train_set = KITTIRawFile(
cfg.root,
cfg.train_file,
with_stereo=cfg.train_stereo,
transform=train_input_transform,
co_transform=co_transform # no target here
)
valid_input_transform = copy.deepcopy(input_transform)
valid_input_transform.transforms.insert(
0, sep_transforms.Zoom(*cfg.test_shape))
valid_set = KITTIFlow(
cfg.flow_data,
with_stereo=cfg.test_stereo,
transform=valid_input_transform,
)
else:
raise NotImplementedError(cfg.type)
return train_set, valid_set
def __init__(self, cfg):
self.cfg = EasyDict(cfg)
self.device = torch.device(
"cuda") if torch.cuda.is_available() else torch.device("cpu")
self.model = self.init_model()
self.input_transform = transforms.Compose([
sep_transforms.Zoom(*self.cfg.test_shape),
sep_transforms.ArrayToTensor(),
transforms.Normalize(mean=[0, 0, 0], std=[255, 255, 255]),
])
def get_dataset(all_cfg):
cfg = all_cfg.data
input_transform = transforms.Compose([
sep_transforms.ArrayToTensor(),
transforms.Normalize(mean=[0, 0, 0], std=[255, 255, 255]),
])
co_transform = get_co_transforms(aug_args=all_cfg.data_aug)
if cfg.type == 'Sintel_Flow':
ap_transform = get_ap_transforms(cfg.at_cfg) if cfg.run_at else None
train_set_1 = Sintel(cfg.root_sintel, n_frames=cfg.train_n_frames, type='clean',
split='training', subsplit=cfg.train_subsplit,
with_flow=False,
ap_transform=ap_transform,
transform=input_transform,
co_transform=co_transform
)
train_set_2 = Sintel(cfg.root_sintel, n_frames=cfg.train_n_frames, type='final',
split='training', subsplit=cfg.train_subsplit,
with_flow=False,
ap_transform=ap_transform,
transform=input_transform,
co_transform=co_transform
)
train_set = ConcatDataset([train_set_1, train_set_2])
valid_input_transform = copy.deepcopy(input_transform)
valid_input_transform.transforms.insert(0, sep_transforms.Zoom(*cfg.test_shape))
valid_set_1 = Sintel(cfg.root_sintel, n_frames=cfg.val_n_frames, type='clean',
split='training', subsplit=cfg.val_subsplit,
transform=valid_input_transform,
target_transform={'flow': sep_transforms.ArrayToTensor()}
)
valid_set_2 = Sintel(cfg.root_sintel, n_frames=cfg.val_n_frames, type='final',
split='training', subsplit=cfg.val_subsplit,
transform=valid_input_transform,
target_transform={'flow': sep_transforms.ArrayToTensor()}
)
valid_set = ConcatDataset([valid_set_1, valid_set_2])
elif cfg.type == 'Sintel_Raw':
train_set = SintelRaw(cfg.root_sintel_raw, n_frames=cfg.train_n_frames,
transform=input_transform, co_transform=co_transform)
valid_input_transform = copy.deepcopy(input_transform)
valid_input_transform.transforms.insert(0, sep_transforms.Zoom(*cfg.test_shape))
valid_set_1 = Sintel(cfg.root_sintel, n_frames=cfg.val_n_frames, type='clean',
split='training', subsplit=cfg.val_subsplit,
transform=valid_input_transform,
target_transform={'flow': sep_transforms.ArrayToTensor()}
)
valid_set_2 = Sintel(cfg.root_sintel, n_frames=cfg.val_n_frames, type='final',
split='training', subsplit=cfg.val_subsplit,
transform=valid_input_transform,
target_transform={'flow': sep_transforms.ArrayToTensor()}
)
valid_set = ConcatDataset([valid_set_1, valid_set_2])
else:
raise NotImplementedError(cfg.type)
return train_set, valid_set