net = module(feature_maps=args.fm,
levels=args.levels,
norm=args.norm,
activation=args.activation,
coi=args.classes_of_interest)
net.load_state_dict(net_state)
break
except RuntimeError:
continue
return net
net = _load_net(args.net, args.device)
if net.__class__ != UNet2D:
net = net.get_unet()
"""
Validate the trained network
"""
validate(net,
test_src.data,
test_src.labels,
args.input_size,
batch_size=args.test_batch_size,
track_progress=True,
write_dir=os.path.join(args.log_dir, 'segmentation'),
val_file=os.path.join(args.log_dir, 'validation.npy'),
device=args.device)
print('[%s] Finished!' % (datetime.datetime.now()))
test_loader_tar_ul,
test_loader_tar_l,
optimizer,
args.epochs,
scheduler=scheduler,
augmenter=augmenter,
print_stats=args.print_stats,
log_dir=args.log_dir,
device=args.device)
"""
Validate the trained network
"""
validate(net.get_unet(),
test_tar_l.data,
test_tar_l.labels,
args.input_size,
batch_size=args.test_batch_size,
write_dir=os.path.join(args.log_dir, 'segmentation_final'),
val_file=os.path.join(args.log_dir, 'validation_final.npy'),
classes_of_interest=args.classes_of_interest)
net.load_state_dict(
torch.load(os.path.join(args.log_dir, 'best_checkpoint.pytorch')))
validate(net.get_unet(),
test_tar_l.data,
test_tar_l.labels,
args.input_size,
batch_size=args.test_batch_size,
write_dir=os.path.join(args.log_dir, 'segmentation_best'),
val_file=os.path.join(args.log_dir, 'validation_best.npy'),
classes_of_interest=args.classes_of_interest)
print('[%s] Finished!' % (datetime.datetime.now()))
loss_fn,
optimizer,
args.epochs,
scheduler=scheduler,
augmenter=augmenter,
print_stats=args.print_stats,
log_dir=args.log_dir,
device=args.device)
"""
Validate the trained network
"""
validate(net,
test.data,
test.labels,
args.input_size,
batch_size=args.test_batch_size,
write_dir=os.path.join(args.log_dir, 'segmentation_final'),
classes_of_interest=args.classes_of_interest,
val_file=os.path.join(args.log_dir, 'validation_final.npy'),
in_channels=args.in_channels)
net = torch.load(os.path.join(args.log_dir, 'best_checkpoint.pytorch'))
validate(net,
test.data,
test.labels,
args.input_size,
batch_size=args.test_batch_size,
write_dir=os.path.join(args.log_dir, 'segmentation_best'),
classes_of_interest=args.classes_of_interest,
val_file=os.path.join(args.log_dir, 'validation_best.npy'),
in_channels=args.in_channels)