CPJKU · superbock · Apr 8, 2016 · Apr 7, 2016
diff --git a/bin/GMMPatternTracker b/bin/GMMPatternTracker
@@ -7,10 +7,8 @@ GMMPatternTracker for tracking (down-)beats based on rhythmic patterns.
 
 from __future__ import absolute_import, division, print_function
 
-import glob
 import argparse
 
-from madmom import MODELS_PATH
 from madmom.processors import IOProcessor, io_arguments
 from madmom.audio.signal import SignalProcessor, FramedSignalProcessor
 from madmom.audio.spectrogram import (FilteredSpectrogramProcessor,
@@ -19,6 +17,7 @@ from madmom.audio.spectrogram import (FilteredSpectrogramProcessor,
                                       MultiBandSpectrogramProcessor)
 from madmom.features import ActivationsProcessor
 from madmom.features.beats import PatternTrackingProcessor
+from madmom.models import PATTERNS_BALLROOM
 
 
 def main():
@@ -87,7 +86,7 @@ def main():
     args.diff_ratio = 0.5
     args.positive_diffs = True
     args.crossover_frequencies = [270]
-    args.pattern_files = glob.glob("%s/patterns/2013/*.pkl" % MODELS_PATH)
+    args.pattern_files = PATTERNS_BALLROOM
 
     # print arguments
     if args.verbose:

diff --git a/bin/SuperFluxNN b/bin/SuperFluxNN
@@ -7,19 +7,18 @@ SuperFlux with neural network based peak picking onset detection algorithm.
 
 from __future__ import absolute_import, division, print_function
 
-import glob
 import argparse
 
-from madmom import MODELS_PATH
 from madmom.processors import IOProcessor, io_arguments
 from madmom.audio.signal import SignalProcessor, FramedSignalProcessor
 from madmom.audio.filters import FilterbankProcessor
 from madmom.audio.spectrogram import (LogarithmicSpectrogramProcessor,
                                       SpectrogramDifferenceProcessor,
                                       SuperFluxProcessor)
-from madmom.ml.nn import NeuralNetworkEnsemble
 from madmom.features import ActivationsProcessor
 from madmom.features.onsets import SpectralOnsetProcessor, PeakPickingProcessor
+from madmom.ml.nn import NeuralNetworkEnsemble
+from madmom.models import ONSETS_BRNN_PP
 
 
 def main():
@@ -82,8 +81,6 @@ def main():
     args.fps = 100
     args.online = False
     args.onset_method = 'superflux'
-    args.nn_files = glob.glob("%s/onsets/2014/onsets_brnn_pp_[1-8].pkl" %
-                              MODELS_PATH)
 
     # print arguments
     if args.verbose:
@@ -107,7 +104,7 @@ def main():
         out_processor = ActivationsProcessor(mode='w', **vars(args))
     else:
         # process everything with multiple RNNs and average the predictions
-        rnn = NeuralNetworkEnsemble.load(args.nn_files)
+        rnn = NeuralNetworkEnsemble.load(ONSETS_BRNN_PP)
         # perform peak picking, i.e. select the local maxima
         pp = PeakPickingProcessor(**vars(args))
         from madmom.utils import write_events as writer

diff --git a/madmom/__init__.py b/madmom/__init__.py
@@ -2,7 +2,7 @@
 # pylint: disable=no-member
 # pylint: disable=wrong-import-position
 """
-Madmom is an audio signal processing library.
+Madmom is an audio and music signal processing library.
 
 This library is used internally by the Department of Computational Perception,
 Johannes Kepler University, Linz, Austria (http://www.cp.jku.at) and the
@@ -15,16 +15,13 @@
 
 from __future__ import absolute_import, division, print_function
 
-import os
 import pkg_resources
 
-MODELS_PATH = '%s/models' % (os.path.dirname(__file__))
+# import all packages
+from . import audio, evaluation, features, ml, models, processors, utils
 
+# define a version variable
 __version__ = pkg_resources.get_distribution("madmom").version
 
 # keep namespace clean
-del os
 del pkg_resources
-
-# finally import all submodules
-from . import audio, features, evaluation, ml, utils, processors
diff --git a/madmom/features/beats.py b/madmom/features/beats.py
@@ -29,12 +29,11 @@ class RNNBeatProcessor(SequentialProcessor):
 
     def __init__(self, post_processor=None, **kwargs):
         # pylint: disable=unused-argument
-        import glob
-        from .. import MODELS_PATH
         from ..audio.signal import SignalProcessor, FramedSignalProcessor
         from ..audio.spectrogram import (
             FilteredSpectrogramProcessor, LogarithmicSpectrogramProcessor,
             SpectrogramDifferenceProcessor)
+        from ..models import BEATS_BLSTM
         from ..ml.nn import NeuralNetworkEnsemble, average_predictions
 
         # define pre-processing chain
@@ -53,13 +52,12 @@ def __init__(self, post_processor=None, **kwargs):
         # stack the features and processes everything sequentially
         pre_processor = SequentialProcessor((sig, multi, np.hstack))
 
-        # process the pre-processed signal with a NN ensemble
-        nn_files = glob.glob("%s/beats/2013/beats_blstm_[1-8].pkl" %
-                             MODELS_PATH)
         # average predictions
         if post_processor is None:
             post_processor = average_predictions
-        nn = NeuralNetworkEnsemble.load(nn_files, ensemble_fn=post_processor)
+        # process the pre-processed signal with a NN ensemble
+        nn = NeuralNetworkEnsemble.load(BEATS_BLSTM,
+                                        ensemble_fn=post_processor)
 
         # instantiate a SequentialProcessor
         super(RNNBeatProcessor, self).__init__((pre_processor, nn))
@@ -302,7 +300,7 @@ class BeatTrackingProcessor(Processor):
     def __init__(self, look_aside=LOOK_ASIDE, look_ahead=LOOK_AHEAD, fps=None,
                  **kwargs):
         # import the TempoEstimation here otherwise we have a loop
-        from madmom.features.tempo import TempoEstimationProcessor
+        from .tempo import TempoEstimationProcessor
         # save variables
         self.look_aside = look_aside
         self.look_ahead = look_ahead
@@ -639,8 +637,7 @@ def add_arguments(parser, interval_sigma=INTERVAL_SIGMA,
 
         """
         # pylint: disable=arguments-differ
-
-        from madmom.utils import OverrideDefaultListAction
+        from ..utils import OverrideDefaultListAction
         # add CRF related arguments
         g = parser.add_argument_group('conditional random field arguments')
         g.add_argument('--interval_sigma', action='store', type=float,
@@ -721,11 +718,10 @@ def __init__(self, min_bpm=MIN_BPM, max_bpm=MAX_BPM, num_tempi=NUM_TEMPI,
                  fps=None, **kwargs):
         # pylint: disable=unused-argument
         # pylint: disable=no-name-in-module
-
-        from madmom.ml.hmm import HiddenMarkovModel as Hmm
         from .beats_hmm import (BeatStateSpace as St,
                                 BeatTransitionModel as Tm,
                                 RNNBeatTrackingObservationModel as Om)
+        from ..ml.hmm import HiddenMarkovModel as Hmm
 
         # convert timing information to construct a beat state space
         min_interval = 60. * fps / max_bpm
@@ -935,11 +931,11 @@ def __init__(self, pattern_files, min_bpm=MIN_BPM, max_bpm=MAX_BPM,
         # pylint: disable=no-name-in-module
 
         import pickle
-        from madmom.ml.hmm import HiddenMarkovModel as Hmm
         from .beats_hmm import (BarStateSpace, BarTransitionModel,
                                 MultiPatternStateSpace,
                                 MultiPatternTransitionModel,
                                 GMMPatternTrackingObservationModel)
+        from ..ml.hmm import HiddenMarkovModel as Hmm
 
         # expand num_tempi and transition_lambda to lists if needed
         if not isinstance(num_tempi, list):
@@ -1066,7 +1062,7 @@ def add_arguments(parser, pattern_files=None, min_bpm=MIN_BPM,
         `transition_lambda` must the same number of items.
 
         """
-        from madmom.utils import OverrideDefaultListAction
+        from ..utils import OverrideDefaultListAction
         # add GMM options
         if pattern_files is not None:
             g = parser.add_argument_group('GMM arguments')

diff --git a/madmom/features/notes.py b/madmom/features/notes.py
@@ -113,9 +113,10 @@ def write_notes(notes, filename, sep='\t', fmt=None, header=''):
     'note_time' 'MIDI_note' ['duration' ['MIDI_velocity']]
 
     """
+    from ..utils import write_events
+    # set default format
     if fmt is None:
         fmt = list(('%.3f', '%d', '%.3f', '%d'))
-    from madmom.utils import write_events
     if not notes.ndim == 2:
         raise ValueError('unknown format for `notes`')
     # truncate to the number of colums given
@@ -153,7 +154,7 @@ def write_midi(notes, filename, duration=0.6, velocity=100):
     'note_time' 'MIDI_note' ['duration' ['MIDI_velocity']]
 
     """
-    from madmom.utils.midi import process_notes
+    from ..utils.midi import process_notes
     # expand the array to have a default duration and velocity
     notes = expand_notes(notes, duration, velocity)
     # write the notes to the file and return them
@@ -189,7 +190,7 @@ def write_mirex_format(notes, filename, duration=0.6):
     'onset_time' 'offset_time' 'note_frequency'
 
     """
-    from madmom.audio.filters import midi2hz
+    from ..audio.filters import midi2hz
     # expand the notes if needed
     notes = expand_notes(notes, duration)
     # report offset time instead of duration
@@ -209,11 +210,11 @@ class RNNPianoNoteProcessor(SequentialProcessor):
 
     def __init__(self, **kwargs):
         # pylint: disable=unused-argument
-        from .. import MODELS_PATH
         from ..audio.signal import SignalProcessor, FramedSignalProcessor
         from ..audio.spectrogram import (
             FilteredSpectrogramProcessor, LogarithmicSpectrogramProcessor,
             SpectrogramDifferenceProcessor)
+        from ..models import NOTES_BRNN
         from ..ml.nn import NeuralNetwork
 
         # define pre-processing chain
@@ -233,7 +234,7 @@ def __init__(self, **kwargs):
         pre_processor = SequentialProcessor((sig, multi, np.hstack))
 
         # process the pre-processed signal with a NN
-        nn = NeuralNetwork.load("%s/notes/2013/notes_brnn.pkl" % MODELS_PATH)
+        nn = NeuralNetwork.load(NOTES_BRNN[0])
 
         # instantiate a SequentialProcessor
         super(RNNPianoNoteProcessor, self).__init__((pre_processor, nn))
diff --git a/madmom/features/onsets.py b/madmom/features/onsets.py
@@ -665,22 +665,19 @@ class RNNOnsetProcessor(SequentialProcessor):
 
     def __init__(self, online=False, **kwargs):
         # pylint: disable=unused-argument
-        import glob
-        from .. import MODELS_PATH
         from ..audio.signal import SignalProcessor, FramedSignalProcessor
         from ..audio.spectrogram import (
             FilteredSpectrogramProcessor, LogarithmicSpectrogramProcessor,
             SpectrogramDifferenceProcessor)
+        from ..models import ONSETS_RNN, ONSETS_BRNN
         from ..ml.nn import NeuralNetworkEnsemble
 
         # choose the appropriate models and set frame sizes accordingly
         if online:
-            nn_files = glob.glob("%s/onsets/2013/onsets_rnn_[1-8].pkl" %
-                                 MODELS_PATH)
+            nn_files = ONSETS_RNN
             frame_sizes = [512, 1024, 2048]
         else:
-            nn_files = glob.glob("%s/onsets/2013/onsets_brnn_[1-8].pkl" %
-                                 MODELS_PATH)
+            nn_files = ONSETS_BRNN
             frame_sizes = [1024, 2048, 4096]
 
         # define pre-processing chain

diff --git a/madmom/models b/madmom/models
diff --git a/tests/__init__.py b/tests/__init__.py
@@ -9,8 +9,6 @@
 
 import os
 
-from madmom import MODELS_PATH
-
 DATA_PATH = os.path.dirname(os.path.realpath(__file__)) + '/data/'
 AUDIO_PATH = DATA_PATH + 'audio/'
 ACTIVATIONS_PATH = DATA_PATH + 'activations/'

diff --git a/tests/test_evaluation_notes.py b/tests/test_evaluation_notes.py
@@ -379,7 +379,6 @@ def setUp(self):
     def test_args(self):
         args = self.parser.parse_args(['notes', ANNOTATIONS_PATH,
                                        DETECTIONS_PATH])
-        print(args)
         self.assertTrue(args.ann_dir is None)
         self.assertTrue(args.ann_suffix == '.notes')
         self.assertTrue(args.det_dir is None)

diff --git a/tests/test_features_beats.py b/tests/test_features_beats.py
@@ -8,11 +8,13 @@
 from __future__ import absolute_import, division, print_function
 
 import unittest
-from . import AUDIO_PATH, ACTIVATIONS_PATH, MODELS_PATH
-from madmom.ml.hmm import HiddenMarkovModel
+from . import AUDIO_PATH, ACTIVATIONS_PATH
 from madmom.features import Activations
 from madmom.features.beats import *
 from madmom.features.beats_hmm import *
+from madmom.ml.hmm import HiddenMarkovModel
+from madmom.models import PATTERNS_BALLROOM
+
 
 sample_file = "%s/sample.wav" % AUDIO_PATH
 sample_act = Activations("%s/sample.beats_blstm_2013.npz" % ACTIVATIONS_PATH)
@@ -96,9 +98,7 @@ def test_process(self):
 class TestPatternTrackingProcessorClass(unittest.TestCase):
 
     def setUp(self):
-        import glob
-        pattern_files = sorted(glob.glob("%s/patterns/2013/*" % MODELS_PATH))
-        self.processor = PatternTrackingProcessor(pattern_files,
+        self.processor = PatternTrackingProcessor(PATTERNS_BALLROOM,
                                                   fps=sample_mb_features.fps)
 
     def test_types(self):

diff --git a/tests/test_features_beats_hmm.py b/tests/test_features_beats_hmm.py
@@ -362,7 +362,6 @@ def test_values_bar(self):
         self.assertTrue(tm.num_states == 20 + 20)
         self.assertTrue(tm.num_transitions == 26 + 28)
         # the first pattern has 26 transitions
-        print(tm.log_probabilities)
         self.assertTrue(np.allclose(tm.states[:26],
                                     [10, 12, 15, 1, 15, 19, 3, 4, 15, 19, 6, 7,
                                      8, 0, 2, 5, 11, 5, 9, 13, 14, 5, 9, 16,

diff --git a/tests/test_features_notes.py b/tests/test_features_notes.py
@@ -26,7 +26,6 @@ class TestLoadNotesFunction(unittest.TestCase):
 
     def test_values(self):
         result = load_notes(ANNOTATIONS_PATH + 'stereo_sample.notes')
-        print(result)
         self.assertTrue(np.allclose(result, NOTES))
 
 
@@ -61,7 +60,6 @@ class TestWriteMirexFormatFunction(unittest.TestCase):
     def test_values(self):
         result = write_mirex_format(NOTES, ANNOTATIONS_PATH +
                                     'stereo_sample.notes.mirex')
-        print(result)
         self.assertTrue(np.allclose(result[:, 0], NOTES[:, 0]))
         self.assertTrue(np.allclose(result[:, 1], NOTES[:, 0] + NOTES[:, 2]))
         self.assertTrue(np.allclose(result[:, 2], [523.3, 87.3, 698.5, 261.6,

diff --git a/tests/test_features_tempo.py b/tests/test_features_tempo.py
@@ -8,7 +8,7 @@
 from __future__ import absolute_import, division, print_function
 
 import unittest
-from . import ACTIVATIONS_PATH, DETECTIONS_PATH
+from . import ACTIVATIONS_PATH
 from madmom.features.tempo import *
 
 act_file = np.load("%s/sample.beats_blstm_2013.npz" % ACTIVATIONS_PATH)
@@ -132,14 +132,12 @@ def test_values(self):
         result = write_tempo(COMB_TEMPI[5], self.out_file)
         self.assertTrue(np.allclose(result, [52.17, 104.34, 1], atol=0.001))
         # multiple tempi given
-        result = write_tempo(COMB_TEMPI, DETECTIONS_PATH +
-                             'sample.tempo_detector.txt')
+        result = write_tempo(COMB_TEMPI, self.out_file)
         self.assertTrue(np.allclose(result, [176.47, 117.65, 0.684],
                                     atol=0.001))
 
     def test_values_mirex(self):
         # multiple tempi given
         result = write_tempo(COMB_TEMPI, self.out_file, mirex=True)
-        print(result)
         self.assertTrue(np.allclose(result, [117.65, 176.47, 0.316],
                                     atol=0.001))