def test_value(self, input_data, expected_value):
percentile = None
if len(input_data) == 3:
[seg_1, seg_2, percentile] = input_data
else:
[seg_1, seg_2] = input_data
ct = 0
seg_1 = torch.tensor(seg_1)
seg_2 = torch.tensor(seg_2)
for metric in ["euclidean", "chessboard", "taxicab"]:
for directed in [True, False]:
hd_metric = HausdorffDistanceMetric(include_background=False,
distance_metric=metric,
percentile=percentile,
directed=directed)
# shape of seg_1, seg_2 are: HWD, converts to BNHWD
batch, n_class = 2, 3
batch_seg_1 = seg_1.unsqueeze(0).unsqueeze(0).repeat(
[batch, n_class, 1, 1, 1])
batch_seg_2 = seg_2.unsqueeze(0).unsqueeze(0).repeat(
[batch, n_class, 1, 1, 1])
hd_metric(batch_seg_1, batch_seg_2)
result = hd_metric.aggregate()
expected_value_curr = expected_value[ct]
np.testing.assert_allclose(expected_value_curr,
result,
rtol=1e-7)
ct += 1
def test_nans(self, input_data):
[seg_1, seg_2] = input_data
seg_1 = torch.tensor(seg_1)
seg_2 = torch.tensor(seg_2)
hd_metric = HausdorffDistanceMetric(include_background=False, get_not_nans=True)
batch_seg_1 = seg_1.unsqueeze(0).unsqueeze(0)
batch_seg_2 = seg_2.unsqueeze(0).unsqueeze(0)
hd_metric(batch_seg_1, batch_seg_2)
result, not_nans = hd_metric.aggregate()
np.testing.assert_allclose(0, result, rtol=1e-7)
np.testing.assert_allclose(0, not_nans, rtol=1e-7)
def __init__(
self,
include_background: bool = False,
distance_metric: str = "euclidean",
percentile: Optional[float] = None,
directed: bool = False,
output_transform: Callable = lambda x: x,
device: Optional[torch.device] = None,
) -> None:
"""
Args:
include_background: whether to include distance computation on the first channel of the predicted output.
Defaults to ``False``.
distance_metric: : [``"euclidean"``, ``"chessboard"``, ``"taxicab"``]
the metric used to compute surface distance. Defaults to ``"euclidean"``.
percentile: an optional float number between 0 and 100. If specified, the corresponding
percentile of the Hausdorff Distance rather than the maximum result will be achieved.
Defaults to ``None``.
directed: whether to calculate directed Hausdorff distance. Defaults to ``False``.
output_transform: transform the ignite.engine.state.output into [y_pred, y] pair.
device: device specification in case of distributed computation usage.
"""
super().__init__(output_transform, device=device)
metric_fn = HausdorffDistanceMetric(
include_background=include_background,
distance_metric=distance_metric,
percentile=percentile,
directed=directed,
reduction=MetricReduction.NONE,
)
super().__init__(metric_fn=metric_fn,
output_transform=output_transform,
device=device)
def __init__(
self,
include_background: bool = False,
distance_metric: str = "euclidean",
percentile: Optional[float] = None,
directed: bool = False,
reduction: Union[MetricReduction, str] = MetricReduction.MEAN,
output_transform: Callable = lambda x: x,
save_details: bool = True,
) -> None:
"""
Args:
include_background: whether to include distance computation on the first channel of the predicted output.
Defaults to ``False``.
distance_metric: : [``"euclidean"``, ``"chessboard"``, ``"taxicab"``]
the metric used to compute surface distance. Defaults to ``"euclidean"``.
percentile: an optional float number between 0 and 100. If specified, the corresponding
percentile of the Hausdorff Distance rather than the maximum result will be achieved.
Defaults to ``None``.
directed: whether to calculate directed Hausdorff distance. Defaults to ``False``.
reduction: {``"none"``, ``"mean"``, ``"sum"``, ``"mean_batch"``, ``"sum_batch"``,
``"mean_channel"``, ``"sum_channel"``}
Define the mode to reduce computation result. Defaults to ``"mean"``.
output_transform: callable to extract `y_pred` and `y` from `ignite.engine.state.output` then
construct `(y_pred, y)` pair, where `y_pred` and `y` can be `batch-first` Tensors or
lists of `channel-first` Tensors. the form of `(y_pred, y)` is required by the `update()`.
`engine.state` and `output_transform` inherit from the ignite concept:
https://pytorch.org/ignite/concepts.html#state, explanation and usage example are in the tutorial:
https://github.com/Project-MONAI/tutorials/blob/master/modules/batch_output_transform.ipynb.
save_details: whether to save metric computation details per image, for example: hausdorff distance
of every image. default to True, will save to `engine.state.metric_details` dict with the metric name as key.
"""
metric_fn = HausdorffDistanceMetric(
include_background=include_background,
distance_metric=distance_metric,
percentile=percentile,
directed=directed,
reduction=reduction,
)
super().__init__(metric_fn=metric_fn,
output_transform=output_transform,
save_details=save_details)
def __init__(
self,
include_background: bool = False,
distance_metric: str = "euclidean",
percentile: Optional[float] = None,
directed: bool = False,
output_transform: Callable = lambda x: x,
save_details: bool = True,
) -> None:
"""
Args:
include_background: whether to include distance computation on the first channel of the predicted output.
Defaults to ``False``.
distance_metric: : [``"euclidean"``, ``"chessboard"``, ``"taxicab"``]
the metric used to compute surface distance. Defaults to ``"euclidean"``.
percentile: an optional float number between 0 and 100. If specified, the corresponding
percentile of the Hausdorff Distance rather than the maximum result will be achieved.
Defaults to ``None``.
directed: whether to calculate directed Hausdorff distance. Defaults to ``False``.
output_transform: callable to extract `y_pred` and `y` from `ignite.engine.state.output` then
construct `(y_pred, y)` pair, where `y_pred` and `y` can be `batch-first` Tensors or
lists of `channel-first` Tensors. the form of `(y_pred, y)` is required by the `update()`.
for example: if `ignite.engine.state.output` is `{"pred": xxx, "label": xxx, "other": xxx}`,
output_transform can be `lambda x: (x["pred"], x["label"])`.
save_details: whether to save metric computation details per image, for example: hausdorff distance
of every image. default to True, will save to `engine.state.metric_details` dict with the metric name as key.
"""
metric_fn = HausdorffDistanceMetric(
include_background=include_background,
distance_metric=distance_metric,
percentile=percentile,
directed=directed,
reduction=MetricReduction.MEAN,
)
super().__init__(
metric_fn=metric_fn,
output_transform=output_transform,
save_details=save_details,
)