# transform = [transforms.ColorJitter(0.5, 0.5, 0.5, 0),
# transforms.RandomAffine(180),
# transforms.RandomErasing(p=1, value=1)]
# train_dataset = ImagenetDataAugDataset(root_dir=args.train_dir, num_wt=3, mask_dim=args.mask_dim, wt=wt,
# filters=filters_cpu, default_transform=default_transform,
# transform=transform, p=0.1)
# Create train dataset
dataset = SampleDataset(file_path=args.sample_file)
data_loader = torch.utils.data.DataLoader(dataset,
batch_size=args.batch_size,
shuffle=True,
num_workers=args.workers,
pin_memory=True,
drop_last=True)
# Model and optimizer
model = UNet_NTail_128_Mod(n_channels=12,
n_classes=3,
n_tails=12,
bilinear=True).to(args.device)
# Load weights
if args.resume:
print('Loading weights')
model = load_weights(model, args.checkpoint_path, args)
eval_biggan_unet128(model, data_loader, args)
transform=default_transform)
valid_loader = torch.utils.data.DataLoader(valid_dataset,
batch_size=args.batch_size,
shuffle=True,
num_workers=args.workers,
pin_memory=True,
drop_last=True)
# Load 128 model
print('Loading model 128 weights')
model_128 = UNet_NTail_128_Mod(n_channels=12,
n_classes=3,
n_tails=12,
bilinear=True).to(args.device)
model_128 = load_weights(model_128, args.model_128_weights, args)
# Model and optimizer
model = UNet_NTail_128_Mod1(n_channels=48,
n_classes=3,
n_tails=48,
bilinear=True).to(args.device)
optimizer = optim.Adam(model.parameters(), lr=args.lr)
state_dict = {'itr': 0}
if args.resume:
print('Loading weights & resuming from iteration {}'.format(
args.checkpoint))
model, optimizer, logger = load_checkpoint(model, optimizer, '256',
args)
train_loader = torch.utils.data.DataLoader(train_dataset,
batch_size=args.batch_size,
shuffle=True,
num_workers=args.workers,
pin_memory=True,
drop_last=True)
# Create validation dataset
valid_dataset = dset.ImageFolder(root=args.valid_dir,
transform=default_transform)
valid_loader = torch.utils.data.DataLoader(valid_dataset,
batch_size=args.batch_size,
shuffle=True,
num_workers=args.workers,
pin_memory=True,
drop_last=True)
# Model and optimizer
model = UNet_NTail_128_Mod(n_channels=48,
n_classes=3,
n_tails=48,
bilinear=True).to(args.device)
# Load weights
print('Loading weights')
model = load_weights(model, args.model_256_weights, args)
eval_unet256(model, train_loader, 'train', args)
eval_unet256(model, valid_loader, 'valid', args)
pin_memory=True,
drop_last=True)
# Create validation dataset
valid_dataset = dset.ImageFolder(root=args.valid_dir,
transform=default_transform)
valid_loader = torch.utils.data.DataLoader(valid_dataset,
batch_size=args.batch_size,
shuffle=True,
num_workers=args.workers,
pin_memory=True,
drop_last=True)
# Model
print('Loading UNet 128 and 256 weights')
model_128 = UNet_NTail_128_Mod(n_channels=12,
n_classes=3,
n_tails=12,
bilinear=True).to(args.device)
model_128 = load_weights(model_128, args.model_128_weights, args)
model_256 = UNet_NTail_128_Mod(n_channels=48,
n_classes=3,
n_tails=48,
bilinear=True).to(args.device)
model_256 = load_weights(model_256, args.model_256_weights, args)
eval_unet_128_256(model_128, model_256, train_loader, 'train', args)
eval_unet_128_256(model_128, model_256, valid_loader, 'valid', args)
