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@0
|
41 FeatureExtractionManager();
|
Chris@0
|
42 virtual ~FeatureExtractionManager();
|
Chris@0
|
43
|
Chris@0
|
44 void setChannels(int channels);
|
Chris@0
|
45 void setDefaultSampleRate(int sampleRate);
|
Chris@0
|
46
|
Chris@0
|
47 bool setSummaryTypes(const set<string> &summaryTypes,
|
Chris@0
|
48 const Vamp::HostExt::PluginSummarisingAdapter::SegmentBoundaries &boundaries);
|
Chris@0
|
49
|
Chris@102
|
50 void setSummariesOnly(bool summariesOnly);
|
Chris@102
|
51
|
Chris@0
|
52 bool addFeatureExtractor(Transform transform,
|
Chris@0
|
53 const vector<FeatureWriter*> &writers);
|
Chris@0
|
54
|
Chris@0
|
55 bool addFeatureExtractorFromFile(QString transformXmlFile,
|
Chris@0
|
56 const vector<FeatureWriter*> &writers);
|
Chris@0
|
57
|
Chris@0
|
58 bool addDefaultFeatureExtractor(TransformId transformId,
|
Chris@0
|
59 const vector<FeatureWriter*> &writers);
|
Chris@0
|
60
|
Chris@47
|
61 // Make a note of an audio or playlist file which will be passed
|
Chris@47
|
62 // to extractFeatures later. Amongst other things, this may
|
Chris@47
|
63 // initialise the default sample rate and channel count
|
Chris@106
|
64 void addSource(QString audioSource, bool willMultiplex);
|
Chris@47
|
65
|
Chris@47
|
66 // Extract features from the given audio or playlist file. If the
|
Chris@47
|
67 // file is a playlist and force is true, continue extracting even
|
Chris@47
|
68 // if a file in the playlist fails.
|
Chris@47
|
69 void extractFeatures(QString audioSource, bool force);
|
Chris@0
|
70
|
Chris@106
|
71 // Extract features from the given audio files, multiplexing into
|
Chris@106
|
72 // a single "file" whose individual channels are mixdowns of the
|
Chris@106
|
73 // supplied sources.
|
Chris@106
|
74 void extractFeaturesMultiplexed(QStringList sources);
|
Chris@106
|
75
|
Chris@0
|
76 private:
|
Chris@0
|
77 // A plugin may have many outputs, so we can have more than one
|
Chris@0
|
78 // transform requested for a single plugin. The things we want to
|
Chris@0
|
79 // run in our process loop are plugins rather than their outputs,
|
Chris@0
|
80 // so we maintain a map from the plugins to the transforms desired
|
Chris@0
|
81 // of them and then iterate through this map
|
Chris@0
|
82
|
Chris@0
|
83 typedef map<Transform, vector<FeatureWriter *> > TransformWriterMap;
|
Chris@0
|
84 typedef map<Vamp::Plugin *, TransformWriterMap> PluginMap;
|
Chris@0
|
85 PluginMap m_plugins;
|
Chris@108
|
86
|
Chris@109
|
87 // When we run plugins, we want to run them in a known order so as
|
Chris@109
|
88 // to get the same results on each run of Sonic Annotator with the
|
Chris@109
|
89 // same transforms. But if we just iterate through our PluginMap,
|
Chris@109
|
90 // we get them in an arbitrary order based on pointer
|
Chris@109
|
91 // address. This vector provides an underlying order for us. Note
|
Chris@109
|
92 // that the TransformWriterMap is consistently ordered (because
|
Chris@109
|
93 // the key is a Transform which has a proper ordering) so using
|
Chris@109
|
94 // this gives us a consistent order across the whole PluginMap
|
Chris@109
|
95 vector<Vamp::Plugin *> m_orderedPlugins;
|
Chris@109
|
96
|
Chris@0
|
97 // And a map back from transforms to their plugins. Note that
|
Chris@0
|
98 // this is keyed by transform, not transform ID -- two differently
|
Chris@0
|
99 // configured transforms with the same ID must use different
|
Chris@0
|
100 // plugin instances.
|
Chris@0
|
101
|
Chris@0
|
102 typedef map<Transform, Vamp::Plugin *> TransformPluginMap;
|
Chris@0
|
103 TransformPluginMap m_transformPluginMap;
|
Chris@0
|
104
|
Chris@0
|
105 // Cache the plugin output descriptors, mapping from plugin to a
|
Chris@0
|
106 // map from output ID to output descriptor.
|
Chris@0
|
107 typedef map<string, Vamp::Plugin::OutputDescriptor> OutputMap;
|
Chris@0
|
108 typedef map<Vamp::Plugin *, OutputMap> PluginOutputMap;
|
Chris@0
|
109 PluginOutputMap m_pluginOutputs;
|
Chris@0
|
110
|
Chris@0
|
111 // Map from plugin output identifier to plugin output index
|
Chris@0
|
112 typedef map<string, int> OutputIndexMap;
|
Chris@0
|
113 OutputIndexMap m_pluginOutputIndices;
|
Chris@0
|
114
|
Chris@0
|
115 typedef set<std::string> SummaryNameSet;
|
Chris@0
|
116 SummaryNameSet m_summaries;
|
Chris@0
|
117 bool m_summariesOnly;
|
Chris@0
|
118 Vamp::HostExt::PluginSummarisingAdapter::SegmentBoundaries m_boundaries;
|
Chris@0
|
119
|
Chris@106
|
120 AudioFileReader *prepareReader(QString audioSource);
|
Chris@106
|
121
|
Chris@106
|
122 void extractFeaturesFor(AudioFileReader *reader, QString audioSource);
|
Chris@106
|
123
|
Chris@0
|
124 void writeSummaries(QString audioSource, Vamp::Plugin *);
|
Chris@0
|
125
|
Chris@0
|
126 void writeFeatures(QString audioSource,
|
Chris@0
|
127 Vamp::Plugin *,
|
Chris@0
|
128 const Vamp::Plugin::FeatureSet &,
|
Chris@0
|
129 Transform::SummaryType summaryType =
|
Chris@0
|
130 Transform::NoSummary);
|
Chris@31
|
131
|
Chris@31
|
132 void testOutputFiles(QString audioSource);
|
Chris@0
|
133 void finish();
|
Chris@0
|
134
|
Chris@0
|
135 int m_blockSize;
|
Chris@0
|
136 int m_defaultSampleRate;
|
Chris@0
|
137 int m_sampleRate;
|
Chris@0
|
138 int m_channels;
|
Chris@45
|
139
|
Chris@45
|
140 QMap<QString, AudioFileReader *> m_readyReaders;
|
Chris@0
|
141
|
Chris@0
|
142 void print(Transform transform) const;
|
Chris@0
|
143 };
|
Chris@0
|
144
|
Chris@0
|
145 #endif
|