add DrumTranscriptor program

Author: Stefan

bin/DrumTranscriptor

@@ -0,0 +1,86 @@
+#!/usr/bin/env python
+# encoding: utf-8
+"""
+Drum transcription with a convolutional recurrent neural network (CRNN).
+
+"""
+
+from __future__ import absolute_import, division, print_function
+
+import argparse
+
+from madmom.features import ActivationsProcessor
+from madmom.features.drums import CRNNDrumProcessor, DrumPeakPickingProcessor
+from madmom.features.notes import write_midi, write_notes
+from madmom.processors import IOProcessor, io_arguments
+
+
+def main():
+    """DrumTranscriptor"""
+
+    # define parser
+    p = argparse.ArgumentParser(
+        formatter_class=argparse.RawDescriptionHelpFormatter, description='''
+    Drum transcription with a convolutional recurrent neural network (CRNN).
+    ''')
+    # version
+    p.add_argument('--version', action='version',
+                   version='DrumTranscriptor.2017')
+    # input/output arguments
+    io_arguments(p, output_suffix='.drum_transcriptor.txt')
+    ActivationsProcessor.add_arguments(p)
+    # peak picking arguments
+    DrumPeakPickingProcessor.add_arguments(
+        p, threshold=0.15, smooth=0, pre_avg=0.1, post_avg=0.01, pre_max=0.02,
+        post_max=0.01, combine=0.02)
+    # midi arguments
+    p.add_argument('--midi', dest='output_format', action='store_const',
+                   const='midi', help='save as MIDI')
+
+    # parse arguments
+    args = p.parse_args()
+
+    # set immutable defaults
+    args.fps = 100
+
+    # set the suffix for midi files
+    if args.output_format == 'midi':
+        args.output_suffix = '.mid'
+
+    # print arguments
+    if args.verbose:
+        print(args)
+
+    # input processor
+    if args.load:
+        # load the activations from file
+        in_processor = ActivationsProcessor(mode='r', **vars(args))
+    else:
+        # use a RNN to predict the notes
+        in_processor = CRNNDrumProcessor(**vars(args))
+
+    # output processor
+    if args.save:
+        # save the RNN note activations to file
+        out_processor = ActivationsProcessor(mode='w', **vars(args))
+    else:
+        # perform peak picking on the activation function
+        peak_picking = DrumPeakPickingProcessor(**vars(args))
+        # output everything in the right format
+        if args.output_format is None:
+            output = write_notes
+        elif args.output_format == 'midi':
+            output = write_midi
+        else:
+            raise ValueError('unknown output format: %s' % args.output_format)
+        out_processor = [peak_picking, output]
+
+    # create an IOProcessor
+    processor = IOProcessor(in_processor, out_processor)
+
+    # and call the processing function
+    args.func(processor, **vars(args))
+
+
+if __name__ == '__main__':
+    main()

tests/data/detections/sample.drum_transcriptor.txt

@@ -0,0 +1,22 @@
+# drums detected by the DrumTranscriptor program with default settings
+0.130	0
+0.130	2
+0.480	2
+0.650	0
+0.800	0
+0.840	1
+0.840	2
+1.010	1
+1.160	0
+1.160	2
+1.500	2
+1.510	0
+1.660	1
+1.660	2
+1.840	0
+1.840	2
+2.180	1
+2.180	2
+2.360	1
+2.520	2
+2.700	0

tests/test_bin.py

@@ -742,6 +742,47 @@ def test_run(self):
         self.assertTrue(np.allclose(result, self.result, atol=1e-5))
 
 
+class TestDrumTranscriptorProgram(unittest.TestCase):
+    def setUp(self):
+        self.bin = pj(program_path, "DrumTranscriptor")
+        self.activations = Activations(
+            pj(ACTIVATIONS_PATH, "sample.drums_crnn.npz"))
+        self.result = np.loadtxt(
+            pj(DETECTIONS_PATH, "sample.drum_transcriptor.txt"))
+
+    def test_help(self):
+        self.assertTrue(run_help(self.bin))
+
+    def test_binary(self):
+        # save activations as binary file
+        run_program([self.bin, '--save', 'single', sample_file,
+                     '-o', tmp_act])
+        act = Activations(tmp_act)
+        self.assertTrue(np.allclose(act, self.activations, atol=1e-5))
+        self.assertEqual(act.fps, self.activations.fps)
+        # reload from file
+        run_program([self.bin, '--load', 'single', tmp_act, '-o', tmp_result])
+        result = np.loadtxt(tmp_result)
+        self.assertTrue(np.allclose(result, self.result, atol=1e-5))
+
+    def test_txt(self):
+        # save activations as txt file
+        run_program([self.bin, '--save', '--sep', ' ', 'single',
+                     sample_file, '-o', tmp_act])
+        act = Activations(tmp_act, sep=' ', fps=100)
+        self.assertTrue(np.allclose(act, self.activations, atol=1e-5))
+        # reload from file
+        run_program([self.bin, '--load', '--sep', ' ', 'single', tmp_act,
+                     '-o', tmp_result])
+        result = np.loadtxt(tmp_result)
+        self.assertTrue(np.allclose(result, self.result, atol=1e-5))
+
+    def test_run(self):
+        run_program([self.bin, 'single', sample_file, '-o', tmp_result])
+        result = np.loadtxt(tmp_result)
+        self.assertTrue(np.allclose(result, self.result, atol=1e-5))
+
+
 class TestSpectralOnsetDetectionProgram(unittest.TestCase):
     def setUp(self):
         self.bin = pj(program_path, "SpectralOnsetDetection")

tests/test_utils.py

@@ -65,6 +65,7 @@
                    pj(DETECTIONS_PATH, 'sample.cnn_onset_detector.txt'),
                    pj(DETECTIONS_PATH, 'sample.complex_flux.txt'),
                    pj(DETECTIONS_PATH, 'sample.crf_beat_detector.txt'),
+                   pj(DETECTIONS_PATH, 'sample.drum_transcriptor.txt'),
                    pj(DETECTIONS_PATH, 'sample.dbn_beat_tracker.txt'),
                    pj(DETECTIONS_PATH, 'sample.dbn_downbeat_tracker.txt'),
                    pj(DETECTIONS_PATH, 'sample.dc_chord_recognition.txt'),