def loss_fx(self):
return DiceLoss(
include_background=False,
to_onehot_y=True,
softmax=True,
reduction=self.reduction,
)
def run_test(batch_size=64, train_steps=100, device=torch.device("cuda:0")):
class _TestBatch(Dataset):
def __getitem__(self, _unused_id):
im, seg = create_test_image_2d(128,
128,
noise_max=1,
num_objs=4,
num_seg_classes=1)
return im[None], seg[None].astype(np.float32)
def __len__(self):
return train_steps
net = UNet(
dimensions=2,
in_channels=1,
out_channels=1,
channels=(4, 8, 16, 32),
strides=(2, 2, 2),
num_res_units=2,
)
loss = DiceLoss(do_sigmoid=True)
opt = torch.optim.Adam(net.parameters(), 1e-4)
src = DataLoader(_TestBatch(), batch_size=batch_size)
trainer = create_supervised_trainer(net, opt, loss, device, False)
trainer.run(src, 1)
loss = trainer.state.output
print('Loss:', loss)
if loss >= 1:
print('Loss value is wrong, expect to be < 1.')
return loss
def test_ill_opts(self):
with self.assertRaisesRegex(ValueError, ""):
MaskedLoss(loss=[])
dice_loss = DiceLoss(include_background=True, sigmoid=True, smooth_nr=1e-5, smooth_dr=1e-5)
with self.assertRaisesRegex(ValueError, ""):
masked = MaskedLoss(loss=dice_loss)
masked(input=torch.zeros((3, 1, 2, 2)), target=torch.zeros((3, 1, 2, 2)), mask=torch.zeros((3, 3, 2, 2)))
with self.assertRaisesRegex(ValueError, ""):
masked = MaskedLoss(loss=dice_loss)
masked(input=torch.zeros((3, 3, 2, 2)), target=torch.zeros((3, 2, 2, 2)), mask=torch.zeros((3, 3, 2, 2)))
def test_shape(self, input_param, input_data, expected_val):
result = DiceLoss(**input_param).forward(**input_data)
self.assertAlmostEqual(result.item(), expected_val, places=5)