Chris@0
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
Chris@0
|
2
|
Chris@0
|
3 /*
|
Chris@0
|
4 Sonic Annotator
|
Chris@0
|
5 A utility for batch feature extraction from audio files.
|
Chris@0
|
6 Mark Levy, Chris Sutton and Chris Cannam, Queen Mary, University of London.
|
Chris@0
|
7 Copyright 2007-2008 QMUL.
|
Chris@0
|
8
|
Chris@0
|
9 This program is free software; you can redistribute it and/or
|
Chris@0
|
10 modify it under the terms of the GNU General Public License as
|
Chris@0
|
11 published by the Free Software Foundation; either version 2 of the
|
Chris@0
|
12 License, or (at your option) any later version. See the file
|
Chris@0
|
13 COPYING included with this distribution for more information.
|
Chris@0
|
14 */
|
Chris@0
|
15
|
Chris@0
|
16 #ifndef _FEATURE_EXTRACTION_MANAGER_H_
|
Chris@0
|
17 #define _FEATURE_EXTRACTION_MANAGER_H_
|
Chris@0
|
18
|
Chris@0
|
19 #include <vector>
|
Chris@0
|
20 #include <set>
|
Chris@0
|
21 #include <string>
|
Chris@0
|
22
|
Chris@45
|
23 #include <QMap>
|
Chris@45
|
24
|
Chris@0
|
25 #include <vamp-hostsdk/Plugin.h>
|
Chris@0
|
26 #include <vamp-hostsdk/PluginSummarisingAdapter.h>
|
Chris@0
|
27 #include <transform/Transform.h>
|
Chris@0
|
28
|
Chris@0
|
29 using std::vector;
|
Chris@0
|
30 using std::set;
|
Chris@0
|
31 using std::string;
|
Chris@0
|
32 using std::pair;
|
Chris@0
|
33 using std::map;
|
Chris@0
|
34
|
Chris@0
|
35 class FeatureWriter;
|
Chris@45
|
36 class AudioFileReader;
|
Chris@0
|
37
|
Chris@0
|
38 class FeatureExtractionManager
|
Chris@0
|
39 {
|
Chris@0
|
40 public:
|
Chris@263
|
41 FeatureExtractionManager(bool verbose);
|
Chris@0
|
42 virtual ~FeatureExtractionManager();
|
Chris@0
|
43
|
Chris@0
|
44 void setChannels(int channels);
|
Chris@0
|
45 void setDefaultSampleRate(int sampleRate);
|
Chris@116
|
46 void setNormalise(bool normalise);
|
Chris@0
|
47
|
Chris@0
|
48 bool setSummaryTypes(const set<string> &summaryTypes,
|
Chris@0
|
49 const Vamp::HostExt::PluginSummarisingAdapter::SegmentBoundaries &boundaries);
|
Chris@0
|
50
|
Chris@102
|
51 void setSummariesOnly(bool summariesOnly);
|
Chris@102
|
52
|
Chris@0
|
53 bool addFeatureExtractor(Transform transform,
|
Chris@0
|
54 const vector<FeatureWriter*> &writers);
|
Chris@0
|
55
|
Chris@0
|
56 bool addFeatureExtractorFromFile(QString transformXmlFile,
|
Chris@0
|
57 const vector<FeatureWriter*> &writers);
|
Chris@0
|
58
|
Chris@0
|
59 bool addDefaultFeatureExtractor(TransformId transformId,
|
Chris@0
|
60 const vector<FeatureWriter*> &writers);
|
Chris@0
|
61
|
Chris@47
|
62 // Make a note of an audio or playlist file which will be passed
|
Chris@47
|
63 // to extractFeatures later. Amongst other things, this may
|
Chris@47
|
64 // initialise the default sample rate and channel count
|
Chris@106
|
65 void addSource(QString audioSource, bool willMultiplex);
|
Chris@47
|
66
|
Chris@47
|
67 // Extract features from the given audio or playlist file. If the
|
Chris@47
|
68 // file is a playlist and force is true, continue extracting even
|
Chris@47
|
69 // if a file in the playlist fails.
|
Chris@112
|
70 void extractFeatures(QString audioSource);
|
Chris@0
|
71
|
Chris@106
|
72 // Extract features from the given audio files, multiplexing into
|
Chris@106
|
73 // a single "file" whose individual channels are mixdowns of the
|
Chris@106
|
74 // supplied sources.
|
Chris@106
|
75 void extractFeaturesMultiplexed(QStringList sources);
|
Chris@106
|
76
|
Chris@0
|
77 private:
|
Chris@263
|
78 bool m_verbose;
|
Chris@263
|
79
|
Chris@0
|
80 // A plugin may have many outputs, so we can have more than one
|
Chris@0
|
81 // transform requested for a single plugin. The things we want to
|
Chris@0
|
82 // run in our process loop are plugins rather than their outputs,
|
Chris@0
|
83 // so we maintain a map from the plugins to the transforms desired
|
Chris@0
|
84 // of them and then iterate through this map
|
Chris@0
|
85
|
Chris@0
|
86 typedef map<Transform, vector<FeatureWriter *> > TransformWriterMap;
|
Chris@0
|
87 typedef map<Vamp::Plugin *, TransformWriterMap> PluginMap;
|
Chris@0
|
88 PluginMap m_plugins;
|
Chris@108
|
89
|
Chris@109
|
90 // When we run plugins, we want to run them in a known order so as
|
Chris@109
|
91 // to get the same results on each run of Sonic Annotator with the
|
Chris@109
|
92 // same transforms. But if we just iterate through our PluginMap,
|
Chris@109
|
93 // we get them in an arbitrary order based on pointer
|
Chris@109
|
94 // address. This vector provides an underlying order for us. Note
|
Chris@109
|
95 // that the TransformWriterMap is consistently ordered (because
|
Chris@109
|
96 // the key is a Transform which has a proper ordering) so using
|
Chris@109
|
97 // this gives us a consistent order across the whole PluginMap
|
Chris@109
|
98 vector<Vamp::Plugin *> m_orderedPlugins;
|
Chris@109
|
99
|
Chris@0
|
100 // And a map back from transforms to their plugins. Note that
|
Chris@0
|
101 // this is keyed by transform, not transform ID -- two differently
|
Chris@0
|
102 // configured transforms with the same ID must use different
|
Chris@0
|
103 // plugin instances.
|
Chris@0
|
104
|
Chris@0
|
105 typedef map<Transform, Vamp::Plugin *> TransformPluginMap;
|
Chris@0
|
106 TransformPluginMap m_transformPluginMap;
|
Chris@0
|
107
|
Chris@0
|
108 // Cache the plugin output descriptors, mapping from plugin to a
|
Chris@0
|
109 // map from output ID to output descriptor.
|
Chris@0
|
110 typedef map<string, Vamp::Plugin::OutputDescriptor> OutputMap;
|
Chris@0
|
111 typedef map<Vamp::Plugin *, OutputMap> PluginOutputMap;
|
Chris@0
|
112 PluginOutputMap m_pluginOutputs;
|
Chris@0
|
113
|
Chris@0
|
114 // Map from plugin output identifier to plugin output index
|
Chris@0
|
115 typedef map<string, int> OutputIndexMap;
|
Chris@0
|
116 OutputIndexMap m_pluginOutputIndices;
|
Chris@0
|
117
|
Chris@0
|
118 typedef set<std::string> SummaryNameSet;
|
Chris@233
|
119 SummaryNameSet m_summaries; // requested on command line for all transforms
|
Chris@233
|
120 bool m_summariesOnly; // command line flag
|
Chris@0
|
121 Vamp::HostExt::PluginSummarisingAdapter::SegmentBoundaries m_boundaries;
|
Chris@0
|
122
|
Chris@106
|
123 AudioFileReader *prepareReader(QString audioSource);
|
Chris@106
|
124
|
Chris@106
|
125 void extractFeaturesFor(AudioFileReader *reader, QString audioSource);
|
Chris@106
|
126
|
Chris@0
|
127 void writeSummaries(QString audioSource, Vamp::Plugin *);
|
Chris@0
|
128
|
Chris@0
|
129 void writeFeatures(QString audioSource,
|
Chris@0
|
130 Vamp::Plugin *,
|
Chris@0
|
131 const Vamp::Plugin::FeatureSet &,
|
Chris@0
|
132 Transform::SummaryType summaryType =
|
Chris@0
|
133 Transform::NoSummary);
|
Chris@31
|
134
|
Chris@31
|
135 void testOutputFiles(QString audioSource);
|
Chris@0
|
136 void finish();
|
Chris@0
|
137
|
Chris@0
|
138 int m_blockSize;
|
Chris@0
|
139 int m_defaultSampleRate;
|
Chris@0
|
140 int m_sampleRate;
|
Chris@0
|
141 int m_channels;
|
Chris@116
|
142 bool m_normalise;
|
Chris@45
|
143
|
Chris@45
|
144 QMap<QString, AudioFileReader *> m_readyReaders;
|
Chris@0
|
145 };
|
Chris@0
|
146
|
Chris@0
|
147 #endif
|