Move test_match_notes, rename no_offset metrics

In addition to moving test_match_notes it has also been updated.

mir_eval/transcription.py

@@ -233,7 +233,7 @@ def precision_recall_f1(ref_intervals, ref_pitches, est_intervals, est_pitches,
     >>> precision, recall, f_measure =
     ...     mir_eval.transcription.precision_recall_f1(ref_intervals,
     ...     ref_pitches, est_intervals, est_pitches)
-    >>> precision_nooffset, recall_nooffset, f_measure_nooffset =
+    >>> precision_no_offset, recall_no_offset, f_measure_no_offset =
     ...     mir_eval.transcription.precision_recall_f1(ref_intervals,
     ...     ref_pitches, est_intervals, est_pitches, offset_ratio=None)
 
@@ -327,23 +327,6 @@ def evaluate(ref_intervals, ref_pitches, est_intervals, est_pitches, **kwargs):
     # Compute all the metrics
     scores = collections.OrderedDict()
 
-    # # Precision, recall and f-measure NOT taking note offsets into account
-    # kwargs['with_offset'] = False
-    # (scores['Precision'],
-    #  scores['Recall'],
-    #  scores['F-measure']) = util.filter_kwargs(precision_recall_f1,
-    #                                            ref_intervals, ref_pitches,
-    #                                            est_intervals, est_pitches,
-    #                                            **kwargs)
-    #
-    # # Precision, recall and f-measure taking note offsets into account
-    # kwargs['with_offset'] = True
-    # (scores['Precision_with_offset'],
-    #  scores['Recall_with_offset'],
-    #  scores['F-measure_with_offset']) = util.filter_kwargs(
-    #     precision_recall_f1, ref_intervals, ref_pitches, est_intervals,
-    #     est_pitches, **kwargs)
-
     # Precision, recall and f-measure taking note offsets into account
     offset_ratio = kwargs.get('offset_ratio', .2)
     if offset_ratio is not None:
@@ -355,9 +338,9 @@ def evaluate(ref_intervals, ref_pitches, est_intervals, est_pitches, **kwargs):
 
     # Precision, recall and f-measure NOT taking note offsets into account
     kwargs['offset_ratio'] = None
-    (scores['Precision_nooffset'],
-     scores['Recall_nooffset'],
-     scores['F-measure_nooffset']) = \
+    (scores['Precision_no_offset'],
+     scores['Recall_no_offset'],
+     scores['F-measure_no_offset']) = \
         util.filter_kwargs(precision_recall_f1, ref_intervals, ref_pitches,
                            est_intervals, est_pitches, **kwargs)
 

tests/test_transcription.py

@@ -1,21 +1,73 @@
 # CREATED: 2/9/16 2:27 PM by Justin Salamon <[email protected]>
 
 import mir_eval
-import json
 import numpy as np
 
 A_TOL = 1e-12
 
 
+def _load_unit_test_reference():
+
+    ref = np.array([
+        [0.100, 0.300, 220.000],
+        [0.300, 0.400, 246.942],
+        [0.500, 0.600, 277.183],
+        [0.550, 0.650, 293.665]])
+
+    return ref[:, :2], ref[:, 2]
+
+
+def _load_unit_test_estimate():
+
+    est = np.array([
+            [0.120,   0.290,   225.000],
+            [0.300,   0.350,   246.942],
+            [0.500,   0.600,   500.000],
+            [0.550,   0.600,   293.665],
+            [0.560,   0.650,   293.665]])
+
+    return est[:, :2], est[:, 2]
+
+
+def _load_unit_test_scores():
+
+    scores = {
+        "Precision": 0.4,
+        "Recall": 0.5,
+        "F-measure": 0.4444444444444445,
+        "Precision_no_offset": 0.6,
+        "Recall_no_offset": 0.75,
+        "F-measure_no_offset": 0.6666666666666665
+    }
+
+    return scores
+
+
+def test_match_notes():
+
+    ref_int, ref_pitch = _load_unit_test_reference()
+    est_int, est_pitch = _load_unit_test_estimate()
+
+    matching = \
+        mir_eval.transcription.match_notes(ref_int, ref_pitch, est_int,
+                                           est_pitch)
+
+    assert matching == [(0, 0), (3, 4)]
+
+    matching = \
+        mir_eval.transcription.match_notes(ref_int, ref_pitch, est_int,
+                                           est_pitch, offset_ratio=None)
+
+    assert matching == [(0, 0), (1, 1), (3, 3)]
+
+
 def test_precision_recall_f1():
 
-    ref_int, ref_pitch = mir_eval.io.load_valued_intervals(
-        'tests/data/transcription/ref00.txt')
-    est_int, est_pitch = mir_eval.io.load_valued_intervals(
-        'tests/data/transcription/est00.txt')
+    # load test data
+    ref_int, ref_pitch = _load_unit_test_reference()
+    est_int, est_pitch = _load_unit_test_estimate()
 
-    # load expected results
-    scores = json.load(open('tests/data/transcription/output00.json', 'rb'))
+    scores = _load_unit_test_scores()
 
     precision, recall, f_measure = \
         mir_eval.transcription.precision_recall_f1(ref_int, ref_pitch, est_int,
@@ -32,7 +84,7 @@ def test_precision_recall_f1():
                                                    offset_ratio=None)
 
     scores_gen = np.array([precision, recall, f_measure])
-    scores_exp = np.array([scores['Precision_nooffset'],
-                           scores['Recall_nooffset'],
-                           scores['F-measure_nooffset']])
+    scores_exp = np.array([scores['Precision_no_offset'],
+                           scores['Recall_no_offset'],
+                           scores['F-measure_no_offset']])
     assert np.allclose(scores_exp, scores_gen, atol=A_TOL)

tests/test_util.py

@@ -158,21 +158,3 @@ def test_bipartite_match():
     for k in matching:
         v = matching[k]
         assert v in G[k] or k in G[v]
-
-
-def test_match_notes():
-
-    ref_int, ref_pitch = mir_eval.io.load_valued_intervals(
-        'tests/data/transcription/ref00.txt')
-    est_int, est_pitch = mir_eval.io.load_valued_intervals(
-        'tests/data/transcription/est00.txt')
-
-    matching = mir_eval.util.match_notes(ref_int, ref_pitch, est_int,
-                                         est_pitch)
-
-    assert matching == [(0, 0), (1, 1), (3, 3)]
-
-    matching = mir_eval.util.match_notes(ref_int, ref_pitch, est_int,
-                                         est_pitch, with_offset=True)
-
-    assert matching == [(0, 0), (3, 4)]