craffel / mir_eval

98%

master: 98%

LAST BUILD ON BRANCH pattern_standard_FPR_edge_case

branch: pattern_standard_FPR_edge_case

CHANGE BRANCH

x

Reset

• pattern_standard_FPR_edge_case
• 0.2
• 0.3
• 0.4
• 0.5
• beat_corner_cases
• bipartite_match_coverage
• changelog
• chord_tuple_index_warning
• continuity_no_one_beat
• coverage
• coveralls-fix
• csr_matrix_toarray
• dependencies
• display
• ejh_20170704_extended_chord_mapping
• filter_kwargs_retain_slop
• getargspec_py35
• hobgoblins
• io-comments
• key
• looser_separation_tolerance
• master
• nonuniform_melody_warning
• open-context
• piano-nopitch-nomidi
• pitch_contour_sonify
• re-pep-8
• remove_boundaries_to_intervals_labels
• remove_evaluators
• remove_rotate_bass_to_root
• remove_unreachable_weighted_accuracy_case
• sonify-clicks-int
• sonify_kwargs
• strict_bass_intervals_arg
• tempo-improvement
• tempo_docs
• tempo_loader
• test_matplotlib_le_3
• transcription
• transcription_vel_no_overlap
• transcriptionext
• use_stringtypes
• velocity_eval

Build # 133

Build Type

push

travis-ci

Committed by craffel

Commit Message

Allow pattern.standard_FPR to take len-1 patterns

When `reference_patterns` or `estimated_patterns` were both of length 1,
then np.diff(P - Q, axis=0) would return an array of shape `(0, 2)`
which would cause a `ValueError` from `np.max`.  This test was basically
just that the change in semitones in the reference and estimated
patterns was the same.  When there's only one note in the pattern, the
change in semitones is the same, so we just add the one-note case as a
special case.

Run Details

1698 of 1730 relevant lines covered (98.15%)

2.94 hits per line