1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
|
"""
Unit tests for mir_eval.segment
"""
import numpy as np
import json
import mir_eval
import glob
import pytest
A_TOL = 1e-12
# Path to the fixture files
REF_GLOB = "data/segment/ref*.lab"
EST_GLOB = "data/segment/est*.lab"
SCORES_GLOB = "data/segment/output*.json"
ref_files = sorted(glob.glob(REF_GLOB))
est_files = sorted(glob.glob(EST_GLOB))
sco_files = sorted(glob.glob(SCORES_GLOB))
assert len(ref_files) == len(est_files) == len(sco_files) > 0
file_sets = list(zip(ref_files, est_files, sco_files))
@pytest.fixture
def segment_data(request):
ref_f, est_f, sco_f = request.param
with open(sco_f) as f:
expected_scores = json.load(f)
# Load in an example segmentation annotation
ref_intervals, ref_labels = mir_eval.io.load_labeled_intervals(ref_f)
# Load in an example segmentation tracker output
est_intervals, est_labels = mir_eval.io.load_labeled_intervals(est_f)
return ref_intervals, ref_labels, est_intervals, est_labels, expected_scores
@pytest.mark.parametrize(
"metric", [mir_eval.segment.detection, mir_eval.segment.deviation]
)
def test_segment_boundary_empty(metric):
with pytest.warns(UserWarning, match="Reference intervals are empty"):
metric(np.zeros((0, 2)), np.array([[1, 2], [2, 3]]), trim=False)
with pytest.warns(UserWarning, match="Estimated intervals are empty"):
metric(np.array([[1, 2], [2, 3]]), np.array([[1, 2]]), trim=True)
with pytest.warns(UserWarning, match="intervals are empty"):
empty_intervals = np.zeros((0, 2))
if metric == mir_eval.segment.detection:
assert np.allclose(metric(empty_intervals, empty_intervals), 0)
else:
assert np.all(np.isnan(metric(empty_intervals, empty_intervals)))
@pytest.mark.xfail(raises=ValueError)
@pytest.mark.parametrize(
"metric", [mir_eval.segment.detection, mir_eval.segment.deviation]
)
@pytest.mark.parametrize(
"intervals",
[
# Now test validation function - intervals must be n by 2
np.array([1, 2, 3, 4]),
# Interval boundaries must be positive
np.array([[-1, 2], [2, 3]]),
# Positive interval durations
np.array([[2, 1], [2, 3]]),
],
)
def test_segment_boundary_errors(metric, intervals):
metric(intervals, intervals)
def test_segment_boundary_detection_perfect():
correct_intervals = np.array([[0, 1], [1, 2]])
assert np.allclose(
mir_eval.segment.detection(correct_intervals, correct_intervals), 1
)
def test_segment_boundary_deviation_perfect():
correct_intervals = np.array([[0, 1], [1, 2]])
assert np.allclose(
mir_eval.segment.deviation(correct_intervals, correct_intervals), 0
)
@pytest.mark.parametrize(
"metric",
[
mir_eval.segment.pairwise,
mir_eval.segment.rand_index,
mir_eval.segment.ari,
mir_eval.segment.mutual_information,
mir_eval.segment.nce,
mir_eval.segment.vmeasure,
],
)
def test_segment_structure_empty(metric):
with pytest.warns(UserWarning, match="Reference intervals are empty"):
metric(np.zeros((0, 2)), [], np.array([[0, 1]]), ["foo"])
with pytest.warns(UserWarning, match="Estimated intervals are empty"):
metric(np.array([[0, 1]]), ["foo"], np.zeros((0, 2)), [])
with pytest.warns(UserWarning, match="intervals are empty"):
empty_intervals = np.zeros((0, 2))
assert np.allclose(metric(empty_intervals, [], empty_intervals, []), 0)
@pytest.mark.xfail(raises=ValueError)
@pytest.mark.parametrize(
"metric",
[
mir_eval.segment.pairwise,
mir_eval.segment.rand_index,
mir_eval.segment.ari,
mir_eval.segment.mutual_information,
mir_eval.segment.nce,
mir_eval.segment.vmeasure,
],
)
@pytest.mark.parametrize(
"intervals, labels",
[
# Test for non-matching numbers of intervals and labels
(np.array([[2, 1], [2, 3]]), ["a", "b", "c"]),
# Now test validation function - intervals must be n by 2
(np.arange(4), ["a", "b", "c", "d"]),
# Interval boundaries must be positive
(np.array([[-1, 2], [2, 3]]), ["a", "b"]),
# Positive interval durations
(np.array([[2, 1], [2, 3]]), ["a", "b"]),
# Number of intervals must match number of labels
(np.array([[2, 1], [2, 3]]), ["a"]),
# Intervals must start at 0
(np.array([[1, 2], [2, 3]]), ["a", "b"]),
],
)
def test_segment_structure_fail(metric, intervals, labels):
metric(intervals, labels, intervals, labels)
@pytest.mark.xfail(raises=ValueError)
@pytest.mark.parametrize(
"metric",
[
mir_eval.segment.pairwise,
mir_eval.segment.rand_index,
mir_eval.segment.ari,
mir_eval.segment.mutual_information,
mir_eval.segment.nce,
mir_eval.segment.vmeasure,
],
)
def test_segment_structure_end_mismatch(metric):
reference_intervals = np.array([[0, 1], [1, 2]])
estimated_intervals = np.array([[0, 1], [1, 3]])
labels = ["a", "b"]
metric(reference_intervals, labels, estimated_intervals, labels)
@pytest.mark.parametrize(
"metric",
[
mir_eval.segment.pairwise,
mir_eval.segment.rand_index,
mir_eval.segment.ari,
mir_eval.segment.mutual_information,
mir_eval.segment.nce,
mir_eval.segment.vmeasure,
],
)
def test_segment_structure_perfect(metric):
reference_intervals = np.array([[0, 1], [1, 2]])
estimated_intervals = np.array([[0, 1], [1, 2]])
labels = ["a", "b"]
if metric == mir_eval.segment.mutual_information:
assert np.allclose(
metric(reference_intervals, labels, estimated_intervals, labels),
[np.log(2), 1, 1],
)
else:
assert np.allclose(
metric(reference_intervals, labels, estimated_intervals, labels), 1
)
@pytest.mark.parametrize("segment_data", file_sets, indirect=True)
def test_segment_functions(segment_data):
ref_intervals, ref_labels, est_intervals, est_labels, expected_scores = segment_data
# Compute scores
scores = mir_eval.segment.evaluate(
ref_intervals, ref_labels, est_intervals, est_labels
)
assert scores.keys() == expected_scores.keys()
for metric in scores:
assert np.allclose(scores[metric], expected_scores[metric], atol=A_TOL)
@pytest.mark.parametrize("segment_data", file_sets, indirect=True)
def test_segment_functions_permuted(segment_data):
ref_intervals, ref_labels, est_intervals, est_labels, expected_scores = segment_data
# Also check with permuted references
idx = np.random.permutation(np.arange(len(ref_intervals)))
perm_int = ref_intervals[idx]
perm_lab = [ref_labels[_] for _ in idx]
scores = mir_eval.segment.evaluate(perm_int, perm_lab, est_intervals, est_labels)
assert scores.keys() == expected_scores.keys()
for metric in scores:
assert np.allclose(scores[metric], expected_scores[metric], atol=A_TOL)
|
