def test_winpr_tensors():
reference_labels = torch.tensor([[0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1]])
computed_labels = torch.tensor([[0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1]])
result = winpr(reference_labels, computed_labels, window_size=3)
assert result["true_positives"] == 4
assert result["true_negatives"] == 40
assert result["false_positives"] == 0
assert result["false_negatives"] == 0
assert result["winP"] == 1
assert result["winR"] == 1
def test_winpr_empty():
reference_labels = [[]]
computed_labels = [[]]
result = winpr(reference_labels, computed_labels, window_size=3)
assert result["true_positives"] == 0
assert result["true_negatives"] == 0
assert result["false_positives"] == 0
assert result["false_negatives"] == 0
assert math.isnan(result["winP"])
assert math.isnan(result["winR"])
def test_winpr_extra_boundaries():
reference_labels = [[0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1]]
computed_labels = [[0, 0, 0, 0, 0, 1, 0, 1, 1, 1, 1, 1]]
result = winpr(reference_labels, computed_labels, window_size=3)
assert result["true_positives"] == 4
assert result["true_negatives"] == 32
assert result["false_positives"] == 8
assert result["false_negatives"] == 0
assert result["winP"] == 1 / 3
assert result["winR"] == 1
def test_winpr_correct():
reference_labels = [[0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1]]
computed_labels = [[0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1]]
result = winpr(reference_labels, computed_labels, window_size=3)
assert result["true_positives"] == 4
assert result["true_negatives"] == 40
assert result["false_positives"] == 0
assert result["false_negatives"] == 0
assert result["winP"] == 1
assert result["winR"] == 1
def test_winpr_missed_boundary():
reference_labels = [[0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1]]
computed_labels = [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]
result = winpr(reference_labels, computed_labels, window_size=3)
assert result["true_positives"] == 0
assert result["true_negatives"] == 40
assert result["false_positives"] == 0
assert result["false_negatives"] == 4
assert result["winP"] is math.nan
assert result["winR"] == 0
def test_winpr_batch_processing_tensors():
real = [
torch.tensor([1, 1, 1]),
torch.tensor([0, 0, 0]),
]
computed = [
torch.tensor([1, 1, 1]),
torch.tensor([0, 0, 0]),
]
result = winpr(real, computed)
assert result["false_positives"] == 0
assert result["false_negatives"] == 0
def test_winpr_near_boundary():
reference_labels = [[0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1]]
computed_labels = [[0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1]]
result = winpr(reference_labels, computed_labels, window_size=3)
assert result["true_positives"] == 3
# TODO: is this correct?
# 40 is the value from the original paper but 39 seems to make more sense to me.
assert result["true_negatives"] == 39
assert result["false_positives"] == 1
assert result["false_negatives"] == 1
assert result["winP"] == 0.75
assert result["winR"] == 0.75