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 #include "FeatureExtractionManager.h"
|
Chris@0
|
17
|
Chris@0
|
18 #include <vamp-hostsdk/PluginChannelAdapter.h>
|
Chris@0
|
19 #include <vamp-hostsdk/PluginBufferingAdapter.h>
|
Chris@0
|
20 #include <vamp-hostsdk/PluginInputDomainAdapter.h>
|
Chris@0
|
21 #include <vamp-hostsdk/PluginSummarisingAdapter.h>
|
Chris@8
|
22 #include <vamp-hostsdk/PluginWrapper.h>
|
Chris@0
|
23 #include <vamp-hostsdk/PluginLoader.h>
|
Chris@0
|
24
|
Chris@0
|
25 #include <iostream>
|
Chris@0
|
26
|
Chris@0
|
27 using namespace std;
|
Chris@0
|
28
|
Chris@0
|
29 using Vamp::Plugin;
|
Chris@0
|
30 using Vamp::PluginBase;
|
Chris@0
|
31 using Vamp::HostExt::PluginLoader;
|
Chris@0
|
32 using Vamp::HostExt::PluginChannelAdapter;
|
Chris@0
|
33 using Vamp::HostExt::PluginBufferingAdapter;
|
Chris@0
|
34 using Vamp::HostExt::PluginInputDomainAdapter;
|
Chris@0
|
35 using Vamp::HostExt::PluginSummarisingAdapter;
|
Chris@8
|
36 using Vamp::HostExt::PluginWrapper;
|
Chris@0
|
37
|
Chris@0
|
38 #include "data/fileio/FileSource.h"
|
Chris@0
|
39 #include "data/fileio/AudioFileReader.h"
|
Chris@0
|
40 #include "data/fileio/AudioFileReaderFactory.h"
|
Chris@0
|
41 #include "data/fileio/PlaylistFileReader.h"
|
Chris@0
|
42 #include "base/TempDirectory.h"
|
Chris@0
|
43 #include "base/ProgressPrinter.h"
|
Chris@0
|
44 #include "transform/TransformFactory.h"
|
Chris@0
|
45 #include "rdf/RDFTransformFactory.h"
|
Chris@0
|
46 #include "transform/FeatureWriter.h"
|
Chris@0
|
47
|
Chris@0
|
48 #include <QTextStream>
|
Chris@0
|
49 #include <QFile>
|
Chris@0
|
50 #include <QFileInfo>
|
Chris@0
|
51
|
Chris@0
|
52 FeatureExtractionManager::FeatureExtractionManager() :
|
Chris@0
|
53 m_summariesOnly(false),
|
Chris@0
|
54 // We can read using an arbitrary fixed block size --
|
Chris@0
|
55 // PluginBufferingAdapter handles this for us. It's likely to be
|
Chris@0
|
56 // quicker to use larger sizes than smallish ones like 1024
|
Chris@0
|
57 m_blockSize(16384),
|
Chris@0
|
58 m_defaultSampleRate(0),
|
Chris@0
|
59 m_sampleRate(0),
|
Chris@0
|
60 m_channels(1)
|
Chris@0
|
61 {
|
Chris@0
|
62 }
|
Chris@0
|
63
|
Chris@0
|
64 FeatureExtractionManager::~FeatureExtractionManager()
|
Chris@0
|
65 {
|
Chris@0
|
66 for (PluginMap::iterator pi = m_plugins.begin();
|
Chris@0
|
67 pi != m_plugins.end(); ++pi) {
|
Chris@0
|
68 delete pi->first;
|
Chris@0
|
69 }
|
Chris@0
|
70 }
|
Chris@0
|
71
|
Chris@0
|
72 void FeatureExtractionManager::setChannels(int channels)
|
Chris@0
|
73 {
|
Chris@0
|
74 m_channels = channels;
|
Chris@0
|
75 }
|
Chris@0
|
76
|
Chris@0
|
77 void FeatureExtractionManager::setDefaultSampleRate(int sampleRate)
|
Chris@0
|
78 {
|
Chris@0
|
79 m_defaultSampleRate = sampleRate;
|
Chris@0
|
80 }
|
Chris@0
|
81
|
Chris@0
|
82 static PluginSummarisingAdapter::SummaryType
|
Chris@0
|
83 getSummaryType(string name)
|
Chris@0
|
84 {
|
Chris@0
|
85 if (name == "min") return PluginSummarisingAdapter::Minimum;
|
Chris@0
|
86 if (name == "max") return PluginSummarisingAdapter::Maximum;
|
Chris@0
|
87 if (name == "mean") return PluginSummarisingAdapter::Mean;
|
Chris@0
|
88 if (name == "median") return PluginSummarisingAdapter::Median;
|
Chris@0
|
89 if (name == "mode") return PluginSummarisingAdapter::Mode;
|
Chris@0
|
90 if (name == "sum") return PluginSummarisingAdapter::Sum;
|
Chris@0
|
91 if (name == "variance") return PluginSummarisingAdapter::Variance;
|
Chris@0
|
92 if (name == "sd") return PluginSummarisingAdapter::StandardDeviation;
|
Chris@0
|
93 if (name == "count") return PluginSummarisingAdapter::Count;
|
Chris@0
|
94 return PluginSummarisingAdapter::UnknownSummaryType;
|
Chris@0
|
95 }
|
Chris@0
|
96
|
Chris@0
|
97 bool FeatureExtractionManager::setSummaryTypes(const set<string> &names,
|
Chris@0
|
98 bool summariesOnly,
|
Chris@0
|
99 const PluginSummarisingAdapter::SegmentBoundaries &boundaries)
|
Chris@0
|
100 {
|
Chris@0
|
101 for (SummaryNameSet::const_iterator i = names.begin();
|
Chris@0
|
102 i != names.end(); ++i) {
|
Chris@0
|
103 if (getSummaryType(*i) == PluginSummarisingAdapter::UnknownSummaryType) {
|
Chris@0
|
104 cerr << "ERROR: Unknown summary type \"" << *i << "\"" << endl;
|
Chris@0
|
105 return false;
|
Chris@0
|
106 }
|
Chris@0
|
107 }
|
Chris@0
|
108 m_summaries = names;
|
Chris@0
|
109 m_summariesOnly = summariesOnly;
|
Chris@0
|
110 m_boundaries = boundaries;
|
Chris@0
|
111 return true;
|
Chris@0
|
112 }
|
Chris@0
|
113
|
Chris@0
|
114 bool FeatureExtractionManager::addFeatureExtractor
|
Chris@0
|
115 (Transform transform, const vector<FeatureWriter*> &writers)
|
Chris@0
|
116 {
|
Chris@0
|
117 //!!! exceptions rather than return values?
|
Chris@0
|
118
|
Chris@0
|
119 if (transform.getSampleRate() == 0) {
|
Chris@0
|
120 if (m_sampleRate == 0) {
|
Chris@0
|
121 cerr << "NOTE: Transform does not specify a sample rate, using default rate of " << m_defaultSampleRate << endl;
|
Chris@0
|
122 transform.setSampleRate(m_defaultSampleRate);
|
Chris@0
|
123 m_sampleRate = m_defaultSampleRate;
|
Chris@0
|
124 } else {
|
Chris@0
|
125 cerr << "NOTE: Transform does not specify a sample rate, using previous transform's rate of " << m_sampleRate << endl;
|
Chris@0
|
126 transform.setSampleRate(m_sampleRate);
|
Chris@0
|
127 }
|
Chris@0
|
128 }
|
Chris@0
|
129
|
Chris@0
|
130 if (m_sampleRate == 0) {
|
Chris@0
|
131 m_sampleRate = transform.getSampleRate();
|
Chris@0
|
132 }
|
Chris@0
|
133
|
Chris@0
|
134 if (transform.getSampleRate() != m_sampleRate) {
|
Chris@0
|
135 cerr << "WARNING: Transform sample rate " << transform.getSampleRate() << " does not match previously specified transform rate of " << m_sampleRate << " -- only a single rate is supported for each run" << endl;
|
Chris@0
|
136 cerr << "WARNING: Using previous rate of " << m_sampleRate << " for this transform as well" << endl;
|
Chris@0
|
137 transform.setSampleRate(m_sampleRate);
|
Chris@0
|
138 }
|
Chris@0
|
139
|
Chris@0
|
140 Plugin *plugin = 0;
|
Chris@0
|
141
|
Chris@0
|
142 // Remember what the original transform looked like, and index
|
Chris@0
|
143 // based on this -- because we may be about to fill in the zeros
|
Chris@0
|
144 // for step and block size, but we want any further copies with
|
Chris@0
|
145 // the same zeros to match this one
|
Chris@0
|
146 Transform originalTransform = transform;
|
Chris@0
|
147
|
Chris@0
|
148 if (m_transformPluginMap.find(transform) == m_transformPluginMap.end()) {
|
Chris@0
|
149
|
Chris@0
|
150 // Test whether we already have a transform that is identical
|
Chris@0
|
151 // to this, except for the output requested and/or the summary
|
Chris@0
|
152 // type -- if so, they should share plugin instances (a vital
|
Chris@0
|
153 // optimisation)
|
Chris@0
|
154
|
Chris@0
|
155 for (TransformPluginMap::iterator i = m_transformPluginMap.begin();
|
Chris@0
|
156 i != m_transformPluginMap.end(); ++i) {
|
Chris@0
|
157 Transform test = i->first;
|
Chris@0
|
158 test.setOutput(transform.getOutput());
|
Chris@0
|
159 test.setSummaryType(transform.getSummaryType());
|
Chris@0
|
160 if (transform == test) {
|
Chris@0
|
161 cerr << "NOTE: Already have transform identical to this one (for \""
|
Chris@0
|
162 << transform.getIdentifier().toStdString()
|
Chris@0
|
163 << "\") in every detail except output identifier and/or "
|
Chris@0
|
164 << "summary type; sharing its plugin instance" << endl;
|
Chris@0
|
165 plugin = i->second;
|
Chris@0
|
166 if (transform.getSummaryType() != Transform::NoSummary &&
|
Chris@0
|
167 !dynamic_cast<PluginSummarisingAdapter *>(plugin)) {
|
Chris@0
|
168 plugin = new PluginSummarisingAdapter(plugin);
|
Chris@0
|
169 i->second = plugin;
|
Chris@0
|
170 }
|
Chris@0
|
171 break;
|
Chris@0
|
172 }
|
Chris@0
|
173 }
|
Chris@0
|
174
|
Chris@0
|
175 if (!plugin) {
|
Chris@0
|
176
|
Chris@0
|
177 TransformFactory *tf = TransformFactory::getInstance();
|
Chris@0
|
178
|
Chris@0
|
179 PluginBase *pb = tf->instantiatePluginFor(transform);
|
Chris@0
|
180 plugin = tf->downcastVampPlugin(pb);
|
Chris@0
|
181 if (!plugin) {
|
Chris@0
|
182 //!!! todo: handle non-Vamp plugins too, or make the main --list
|
Chris@0
|
183 // option print out only Vamp transforms
|
Chris@0
|
184 cerr << "ERROR: Failed to load plugin for transform \""
|
Chris@0
|
185 << transform.getIdentifier().toStdString() << "\"" << endl;
|
Chris@0
|
186 delete pb;
|
Chris@0
|
187 return false;
|
Chris@0
|
188 }
|
Chris@0
|
189
|
Chris@0
|
190 // We will provide the plugin with arbitrary step and
|
Chris@0
|
191 // block sizes (so that we can use the same read/write
|
Chris@0
|
192 // block size for all transforms), and to that end we use
|
Chris@0
|
193 // a PluginBufferingAdapter. However, we need to know the
|
Chris@0
|
194 // underlying step size so that we can provide the right
|
Chris@0
|
195 // context for dense outputs. (Although, don't forget
|
Chris@0
|
196 // that the PluginBufferingAdapter rewrites
|
Chris@0
|
197 // OneSamplePerStep outputs so as to use FixedSampleRate
|
Chris@0
|
198 // -- so it supplies the sample rate in the output
|
Chris@0
|
199 // feature. I'm not sure whether we can easily use that.)
|
Chris@0
|
200
|
Chris@0
|
201 size_t pluginStepSize = plugin->getPreferredStepSize();
|
Chris@0
|
202 size_t pluginBlockSize = plugin->getPreferredBlockSize();
|
Chris@0
|
203
|
Chris@0
|
204 // adapt the plugin for buffering, channels, etc.
|
Chris@0
|
205 if (plugin->getInputDomain() == Plugin::FrequencyDomain) {
|
Chris@0
|
206 plugin = new PluginInputDomainAdapter(plugin);
|
Chris@0
|
207 }
|
Chris@0
|
208
|
Chris@0
|
209 PluginBufferingAdapter *pba = new PluginBufferingAdapter(plugin);
|
Chris@0
|
210 plugin = pba;
|
Chris@0
|
211
|
Chris@0
|
212 if (transform.getStepSize() != 0) {
|
Chris@0
|
213 pba->setPluginStepSize(transform.getStepSize());
|
Chris@0
|
214 } else {
|
Chris@0
|
215 transform.setStepSize(pluginStepSize);
|
Chris@0
|
216 }
|
Chris@0
|
217
|
Chris@0
|
218 if (transform.getBlockSize() != 0) {
|
Chris@0
|
219 pba->setPluginBlockSize(transform.getBlockSize());
|
Chris@0
|
220 } else {
|
Chris@0
|
221 transform.setBlockSize(pluginBlockSize);
|
Chris@0
|
222 }
|
Chris@0
|
223
|
Chris@0
|
224 plugin = new PluginChannelAdapter(plugin);
|
Chris@0
|
225
|
Chris@0
|
226 if (!m_summaries.empty() ||
|
Chris@0
|
227 transform.getSummaryType() != Transform::NoSummary) {
|
Chris@0
|
228 PluginSummarisingAdapter *adapter =
|
Chris@0
|
229 new PluginSummarisingAdapter(plugin);
|
Chris@0
|
230 adapter->setSummarySegmentBoundaries(m_boundaries);
|
Chris@0
|
231 plugin = adapter;
|
Chris@0
|
232 }
|
Chris@0
|
233
|
Chris@0
|
234 if (!plugin->initialise(m_channels, m_blockSize, m_blockSize)) {
|
Chris@0
|
235 cerr << "ERROR: Plugin initialise (channels = " << m_channels << ", stepSize = " << m_blockSize << ", blockSize = " << m_blockSize << ") failed." << endl;
|
Chris@0
|
236 delete plugin;
|
Chris@0
|
237 return false;
|
Chris@0
|
238 }
|
Chris@0
|
239
|
Chris@0
|
240 // cerr << "Initialised plugin" << endl;
|
Chris@0
|
241
|
Chris@0
|
242 size_t actualStepSize = 0;
|
Chris@0
|
243 size_t actualBlockSize = 0;
|
Chris@0
|
244 pba->getActualStepAndBlockSizes(actualStepSize, actualBlockSize);
|
Chris@0
|
245 transform.setStepSize(actualStepSize);
|
Chris@0
|
246 transform.setBlockSize(actualBlockSize);
|
Chris@0
|
247
|
Chris@0
|
248 Plugin::OutputList outputs = plugin->getOutputDescriptors();
|
Chris@0
|
249 for (int i = 0; i < (int)outputs.size(); ++i) {
|
Chris@0
|
250
|
Chris@0
|
251 // cerr << "Newly initialised plugin output " << i << " has bin count " << outputs[i].binCount << endl;
|
Chris@0
|
252
|
Chris@0
|
253 m_pluginOutputs[plugin][outputs[i].identifier] = outputs[i];
|
Chris@0
|
254 m_pluginOutputIndices[outputs[i].identifier] = i;
|
Chris@0
|
255 }
|
Chris@0
|
256
|
Chris@10
|
257 cerr << "NOTE: Loaded and initialised plugin for transform \""
|
Chris@0
|
258 << transform.getIdentifier().toStdString() << "\"" << endl;
|
Chris@8
|
259
|
Chris@8
|
260 } else {
|
Chris@8
|
261
|
Chris@8
|
262 if (transform.getStepSize() == 0 || transform.getBlockSize() == 0) {
|
Chris@8
|
263
|
Chris@8
|
264 PluginWrapper *pw = dynamic_cast<PluginWrapper *>(plugin);
|
Chris@8
|
265 if (pw) {
|
Chris@8
|
266 PluginBufferingAdapter *pba =
|
Chris@8
|
267 pw->getWrapper<PluginBufferingAdapter>();
|
Chris@8
|
268 if (pba) {
|
Chris@8
|
269 size_t actualStepSize = 0;
|
Chris@8
|
270 size_t actualBlockSize = 0;
|
Chris@8
|
271 pba->getActualStepAndBlockSizes(actualStepSize,
|
Chris@8
|
272 actualBlockSize);
|
Chris@8
|
273 if (transform.getStepSize() == 0) {
|
Chris@8
|
274 transform.setStepSize(actualStepSize);
|
Chris@8
|
275 }
|
Chris@8
|
276 if (transform.getBlockSize() == 0) {
|
Chris@8
|
277 transform.setBlockSize(actualBlockSize);
|
Chris@8
|
278 }
|
Chris@8
|
279 }
|
Chris@8
|
280 }
|
Chris@8
|
281 }
|
Chris@0
|
282 }
|
Chris@0
|
283
|
Chris@0
|
284 if (transform.getOutput() == "") {
|
Chris@0
|
285 transform.setOutput
|
Chris@0
|
286 (plugin->getOutputDescriptors()[0].identifier.c_str());
|
Chris@0
|
287 }
|
Chris@0
|
288
|
Chris@0
|
289 m_transformPluginMap[transform] = plugin;
|
Chris@0
|
290
|
Chris@0
|
291 if (!(originalTransform == transform)) {
|
Chris@0
|
292 m_transformPluginMap[originalTransform] = plugin;
|
Chris@0
|
293 }
|
Chris@0
|
294
|
Chris@0
|
295 } else {
|
Chris@0
|
296
|
Chris@0
|
297 plugin = m_transformPluginMap[transform];
|
Chris@0
|
298 }
|
Chris@0
|
299
|
Chris@0
|
300 m_plugins[plugin][transform] = writers;
|
Chris@0
|
301
|
Chris@0
|
302 return true;
|
Chris@0
|
303 }
|
Chris@0
|
304
|
Chris@0
|
305 bool FeatureExtractionManager::addDefaultFeatureExtractor
|
Chris@0
|
306 (TransformId transformId, const vector<FeatureWriter*> &writers)
|
Chris@0
|
307 {
|
Chris@0
|
308 TransformFactory *tf = TransformFactory::getInstance();
|
Chris@0
|
309
|
Chris@0
|
310 if (m_sampleRate == 0) {
|
Chris@0
|
311 if (m_defaultSampleRate == 0) {
|
Chris@0
|
312 cerr << "ERROR: Default transform requested, but no default sample rate available" << endl;
|
Chris@0
|
313 return false;
|
Chris@0
|
314 } else {
|
Chris@0
|
315 cerr << "NOTE: Using default sample rate of " << m_defaultSampleRate << " for default transform" << endl;
|
Chris@0
|
316 m_sampleRate = m_defaultSampleRate;
|
Chris@0
|
317 }
|
Chris@0
|
318 }
|
Chris@0
|
319
|
Chris@0
|
320 Transform transform = tf->getDefaultTransformFor(transformId, m_sampleRate);
|
Chris@0
|
321
|
Chris@0
|
322 return addFeatureExtractor(transform, writers);
|
Chris@0
|
323 }
|
Chris@0
|
324
|
Chris@0
|
325 bool FeatureExtractionManager::addFeatureExtractorFromFile
|
Chris@0
|
326 (QString transformXmlFile, const vector<FeatureWriter*> &writers)
|
Chris@0
|
327 {
|
Chris@0
|
328 RDFTransformFactory factory
|
Chris@0
|
329 (QUrl::fromLocalFile(QFileInfo(transformXmlFile).absoluteFilePath())
|
Chris@0
|
330 .toString());
|
Chris@0
|
331 ProgressPrinter printer("Parsing transforms RDF file");
|
Chris@0
|
332 std::vector<Transform> transforms = factory.getTransforms(&printer);
|
Chris@0
|
333 if (!factory.isOK()) {
|
Chris@0
|
334 cerr << "WARNING: FeatureExtractionManager::addFeatureExtractorFromFile: Failed to parse transforms file: " << factory.getErrorString().toStdString() << endl;
|
Chris@0
|
335 if (factory.isRDF()) {
|
Chris@0
|
336 return false; // no point trying it as XML
|
Chris@0
|
337 }
|
Chris@0
|
338 }
|
Chris@0
|
339 if (!transforms.empty()) {
|
Chris@0
|
340 bool success = true;
|
Chris@0
|
341 for (int i = 0; i < (int)transforms.size(); ++i) {
|
Chris@0
|
342 if (!addFeatureExtractor(transforms[i], writers)) {
|
Chris@0
|
343 success = false;
|
Chris@0
|
344 }
|
Chris@0
|
345 }
|
Chris@0
|
346 return success;
|
Chris@0
|
347 }
|
Chris@0
|
348
|
Chris@0
|
349 QFile file(transformXmlFile);
|
Chris@0
|
350 if (!file.open(QIODevice::ReadOnly | QIODevice::Text)) {
|
Chris@0
|
351 cerr << "ERROR: Failed to open transform XML file \""
|
Chris@0
|
352 << transformXmlFile.toStdString() << "\" for reading" << endl;
|
Chris@0
|
353 return false;
|
Chris@0
|
354 }
|
Chris@0
|
355
|
Chris@0
|
356 QTextStream *qts = new QTextStream(&file);
|
Chris@0
|
357 QString qs = qts->readAll();
|
Chris@0
|
358 delete qts;
|
Chris@0
|
359 file.close();
|
Chris@0
|
360
|
Chris@0
|
361 Transform transform(qs);
|
Chris@0
|
362
|
Chris@0
|
363 return addFeatureExtractor(transform, writers);
|
Chris@0
|
364 }
|
Chris@0
|
365
|
Chris@0
|
366 void FeatureExtractionManager::extractFeatures(QString audioSource)
|
Chris@0
|
367 {
|
Chris@0
|
368 if (m_plugins.empty()) return;
|
Chris@0
|
369
|
Chris@11
|
370 ProgressPrinter retrievalProgress("Retrieving audio data...");
|
Chris@0
|
371
|
Chris@11
|
372 FileSource source(audioSource, &retrievalProgress);
|
Chris@0
|
373 if (!source.isAvailable()) {
|
Chris@0
|
374 cerr << "ERROR: File or URL \"" << audioSource.toStdString()
|
Chris@0
|
375 << "\" could not be located" << endl;
|
Chris@0
|
376 exit(1);
|
Chris@0
|
377 }
|
Chris@0
|
378
|
Chris@0
|
379 source.waitForData();
|
Chris@0
|
380
|
Chris@0
|
381 if (QFileInfo(audioSource).suffix().toLower() == "m3u") {
|
Chris@0
|
382 PlaylistFileReader reader(source);
|
Chris@0
|
383 if (reader.isOK()) {
|
Chris@0
|
384 vector<QString> files = reader.load();
|
Chris@0
|
385 for (int i = 0; i < (int)files.size(); ++i) {
|
Chris@0
|
386 extractFeatures(files[i]);
|
Chris@0
|
387 }
|
Chris@0
|
388 return;
|
Chris@0
|
389 } else {
|
Chris@0
|
390 cerr << "ERROR: Playlist \"" << audioSource.toStdString()
|
Chris@0
|
391 << "\" could not be opened" << endl;
|
Chris@0
|
392 exit(1);
|
Chris@0
|
393 }
|
Chris@0
|
394 }
|
Chris@0
|
395
|
Chris@0
|
396 if (m_sampleRate == 0) {
|
Chris@0
|
397 cerr << "ERROR: Internal error in FeatureExtractionManager::extractFeatures: Plugin list is non-empty, but no sample rate set" << endl;
|
Chris@0
|
398 exit(1);
|
Chris@0
|
399 }
|
Chris@0
|
400
|
Chris@0
|
401 AudioFileReader *reader =
|
Chris@11
|
402 AudioFileReaderFactory::createReader(source, m_sampleRate, &retrievalProgress);
|
Chris@0
|
403
|
Chris@0
|
404 if (!reader) {
|
Chris@0
|
405 cerr << "ERROR: File or URL \"" << audioSource.toStdString()
|
Chris@0
|
406 << "\" could not be opened" << endl;
|
Chris@0
|
407 exit(1);
|
Chris@0
|
408 }
|
Chris@0
|
409
|
Chris@0
|
410 size_t channels = reader->getChannelCount();
|
Chris@11
|
411
|
Chris@11
|
412 retrievalProgress.done();
|
Chris@11
|
413
|
Chris@0
|
414 cerr << "Opened " << channels << "-channel file or URL \"" << audioSource.toStdString() << "\"" << endl;
|
Chris@0
|
415
|
Chris@0
|
416 // reject file if it has too few channels, plugin will handle if it has too many
|
Chris@0
|
417 if ((int)channels < m_channels) {
|
Chris@0
|
418 //!!! should not be terminating here!
|
Chris@0
|
419 cerr << "ERROR: File or URL \"" << audioSource.toStdString() << "\" has less than " << m_channels << " channels" << endl;
|
Chris@0
|
420 exit(1);
|
Chris@0
|
421 }
|
Chris@0
|
422
|
Chris@0
|
423 // allocate audio buffers
|
Chris@0
|
424 float **data = new float *[m_channels];
|
Chris@0
|
425 for (int c = 0; c < m_channels; ++c) {
|
Chris@0
|
426 data[c] = new float[m_blockSize];
|
Chris@0
|
427 }
|
Chris@0
|
428
|
Chris@0
|
429 size_t frameCount = reader->getFrameCount();
|
Chris@0
|
430
|
Chris@0
|
431 // cerr << "file has " << frameCount << " frames" << endl;
|
Chris@0
|
432
|
Chris@0
|
433 for (PluginMap::iterator pi = m_plugins.begin();
|
Chris@0
|
434 pi != m_plugins.end(); ++pi) {
|
Chris@0
|
435
|
Chris@0
|
436 Plugin *plugin = pi->first;
|
Chris@0
|
437
|
Chris@0
|
438 // std::cerr << "Calling reset on " << plugin << std::endl;
|
Chris@0
|
439 plugin->reset();
|
Chris@0
|
440
|
Chris@0
|
441 for (TransformWriterMap::iterator ti = pi->second.begin();
|
Chris@0
|
442 ti != pi->second.end(); ++ti) {
|
Chris@0
|
443
|
Chris@0
|
444 const Transform &transform = ti->first;
|
Chris@0
|
445
|
Chris@0
|
446 //!!! we may want to set the start and duration times for extraction
|
Chris@0
|
447 // in the transform record (defaults of zero indicate extraction
|
Chris@0
|
448 // from the whole file)
|
Chris@0
|
449 // transform.setStartTime(RealTime::zeroTime);
|
Chris@0
|
450 // transform.setDuration
|
Chris@0
|
451 // (RealTime::frame2RealTime(reader->getFrameCount(), m_sampleRate));
|
Chris@0
|
452
|
Chris@0
|
453 string outputId = transform.getOutput().toStdString();
|
Chris@0
|
454 if (m_pluginOutputs[plugin].find(outputId) ==
|
Chris@0
|
455 m_pluginOutputs[plugin].end()) {
|
Chris@0
|
456 //!!! throw?
|
Chris@0
|
457 cerr << "WARNING: Nonexistent plugin output \"" << outputId << "\" requested for transform \""
|
Chris@0
|
458 << transform.getIdentifier().toStdString() << "\", ignoring this transform"
|
Chris@0
|
459 << endl;
|
Chris@0
|
460 /*
|
Chris@0
|
461 cerr << "Known outputs for all plugins are as follows:" << endl;
|
Chris@0
|
462 for (PluginOutputMap::const_iterator k = m_pluginOutputs.begin();
|
Chris@0
|
463 k != m_pluginOutputs.end(); ++k) {
|
Chris@0
|
464 cerr << "Plugin " << k->first << ": ";
|
Chris@0
|
465 if (k->second.empty()) {
|
Chris@0
|
466 cerr << "(none)";
|
Chris@0
|
467 }
|
Chris@0
|
468 for (OutputMap::const_iterator i = k->second.begin();
|
Chris@0
|
469 i != k->second.end(); ++i) {
|
Chris@0
|
470 cerr << "\"" << i->first << "\" ";
|
Chris@0
|
471 }
|
Chris@0
|
472 cerr << endl;
|
Chris@0
|
473 }
|
Chris@0
|
474 */
|
Chris@0
|
475 }
|
Chris@0
|
476 }
|
Chris@0
|
477 }
|
Chris@0
|
478
|
Chris@0
|
479 long startFrame = 0;
|
Chris@0
|
480 long endFrame = frameCount;
|
Chris@0
|
481
|
Chris@0
|
482 /*!!! No -- there is no single transform to pull this stuff from --
|
Chris@0
|
483 * the transforms may have various start and end times, need to be far
|
Chris@0
|
484 * cleverer about this if we're going to support them
|
Chris@0
|
485
|
Chris@0
|
486 RealTime trStartRT = transform.getStartTime();
|
Chris@0
|
487 RealTime trDurationRT = transform.getDuration();
|
Chris@0
|
488
|
Chris@0
|
489 long trStart = RealTime::realTime2Frame(trStartRT, m_sampleRate);
|
Chris@0
|
490 long trDuration = RealTime::realTime2Frame(trDurationRT, m_sampleRate);
|
Chris@0
|
491
|
Chris@0
|
492 if (trStart == 0 || trStart < startFrame) {
|
Chris@0
|
493 trStart = startFrame;
|
Chris@0
|
494 }
|
Chris@0
|
495
|
Chris@0
|
496 if (trDuration == 0) {
|
Chris@0
|
497 trDuration = endFrame - trStart;
|
Chris@0
|
498 }
|
Chris@0
|
499 if (trStart + trDuration > endFrame) {
|
Chris@0
|
500 trDuration = endFrame - trStart;
|
Chris@0
|
501 }
|
Chris@0
|
502
|
Chris@0
|
503 startFrame = trStart;
|
Chris@0
|
504 endFrame = trStart + trDuration;
|
Chris@0
|
505 */
|
Chris@0
|
506
|
Chris@0
|
507 for (PluginMap::iterator pi = m_plugins.begin();
|
Chris@0
|
508 pi != m_plugins.end(); ++pi) {
|
Chris@0
|
509
|
Chris@0
|
510 for (TransformWriterMap::const_iterator ti = pi->second.begin();
|
Chris@0
|
511 ti != pi->second.end(); ++ti) {
|
Chris@0
|
512
|
Chris@0
|
513 const vector<FeatureWriter *> &writers = ti->second;
|
Chris@0
|
514
|
Chris@0
|
515 for (int j = 0; j < (int)writers.size(); ++j) {
|
Chris@0
|
516 FeatureWriter::TrackMetadata m;
|
Chris@0
|
517 m.title = reader->getTitle();
|
Chris@0
|
518 m.maker = reader->getMaker();
|
Chris@0
|
519 writers[j]->setTrackMetadata(audioSource, m);
|
Chris@0
|
520 }
|
Chris@0
|
521 }
|
Chris@0
|
522 }
|
Chris@0
|
523
|
Chris@0
|
524 ProgressPrinter extractionProgress("Extracting and writing features...");
|
Chris@0
|
525 int progress = 0;
|
Chris@0
|
526
|
Chris@0
|
527 for (long i = startFrame; i < endFrame; i += m_blockSize) {
|
Chris@0
|
528
|
Chris@0
|
529 //!!! inefficient, although much of the inefficiency may be
|
Chris@0
|
530 // susceptible to optimisation
|
Chris@0
|
531
|
Chris@0
|
532 SampleBlock frames;
|
Chris@0
|
533 reader->getInterleavedFrames(i, m_blockSize, frames);
|
Chris@0
|
534
|
Chris@0
|
535 // We have to do our own channel handling here; we can't just
|
Chris@0
|
536 // leave it to the plugin adapter because the same plugin
|
Chris@0
|
537 // adapter may have to serve for input files with various
|
Chris@0
|
538 // numbers of channels (so the adapter is simply configured
|
Chris@0
|
539 // with a fixed channel count, generally 1).
|
Chris@0
|
540
|
Chris@0
|
541 int rc = reader->getChannelCount();
|
Chris@0
|
542
|
Chris@0
|
543 for (int j = 0; j < m_blockSize; ++j) {
|
Chris@0
|
544 for (int c = 0; c < m_channels; ++c) {
|
Chris@0
|
545 int index;
|
Chris@0
|
546 if (c < rc) {
|
Chris@0
|
547 index = j * rc + c;
|
Chris@0
|
548 data[c][j] = 0.f;
|
Chris@0
|
549 } else {
|
Chris@0
|
550 index = j * rc + (c % rc);
|
Chris@0
|
551 }
|
Chris@0
|
552 if (index < (int)frames.size()) {
|
Chris@0
|
553 data[c][j] += frames[index];
|
Chris@0
|
554 }
|
Chris@0
|
555 }
|
Chris@0
|
556 }
|
Chris@0
|
557
|
Chris@0
|
558 Vamp::RealTime timestamp = Vamp::RealTime::frame2RealTime
|
Chris@0
|
559 (i, m_sampleRate);
|
Chris@0
|
560
|
Chris@0
|
561 for (PluginMap::iterator pi = m_plugins.begin();
|
Chris@0
|
562 pi != m_plugins.end(); ++pi) {
|
Chris@0
|
563
|
Chris@0
|
564 Plugin *plugin = pi->first;
|
Chris@0
|
565 Plugin::FeatureSet featureSet = plugin->process(data, timestamp);
|
Chris@0
|
566
|
Chris@0
|
567 if (!m_summariesOnly) {
|
Chris@0
|
568 writeFeatures(audioSource, plugin, featureSet);
|
Chris@0
|
569 }
|
Chris@0
|
570 }
|
Chris@0
|
571
|
Chris@0
|
572 int pp = progress;
|
Chris@6
|
573 progress = int(((i - startFrame) * 100.0) / (endFrame - startFrame) + 0.1);
|
Chris@0
|
574 if (progress > pp) extractionProgress.setProgress(progress);
|
Chris@0
|
575 }
|
Chris@10
|
576
|
Chris@12
|
577 std::cerr << "FeatureExtractionManager: deleting audio file reader" << std::endl;
|
Chris@12
|
578
|
Chris@10
|
579 delete reader;
|
Chris@0
|
580
|
Chris@0
|
581 for (PluginMap::iterator pi = m_plugins.begin();
|
Chris@0
|
582 pi != m_plugins.end(); ++pi) {
|
Chris@0
|
583
|
Chris@0
|
584 Plugin *plugin = pi->first;
|
Chris@0
|
585 Plugin::FeatureSet featureSet = plugin->getRemainingFeatures();
|
Chris@0
|
586
|
Chris@0
|
587 if (!m_summariesOnly) {
|
Chris@0
|
588 writeFeatures(audioSource, plugin, featureSet);
|
Chris@0
|
589 }
|
Chris@0
|
590
|
Chris@0
|
591 if (!m_summaries.empty()) {
|
Chris@0
|
592 PluginSummarisingAdapter *adapter =
|
Chris@0
|
593 dynamic_cast<PluginSummarisingAdapter *>(plugin);
|
Chris@0
|
594 if (!adapter) {
|
Chris@0
|
595 cerr << "WARNING: Summaries requested, but plugin is not a summarising adapter" << endl;
|
Chris@0
|
596 } else {
|
Chris@0
|
597 for (SummaryNameSet::const_iterator sni = m_summaries.begin();
|
Chris@0
|
598 sni != m_summaries.end(); ++sni) {
|
Chris@0
|
599 featureSet.clear();
|
Chris@0
|
600 //!!! problem here -- we are requesting summaries
|
Chris@0
|
601 //!!! for all outputs, but they in principle have
|
Chris@0
|
602 //!!! different averaging requirements depending
|
Chris@0
|
603 //!!! on whether their features have duration or
|
Chris@0
|
604 //!!! not
|
Chris@0
|
605 featureSet = adapter->getSummaryForAllOutputs
|
Chris@0
|
606 (getSummaryType(*sni),
|
Chris@0
|
607 PluginSummarisingAdapter::ContinuousTimeAverage);
|
Chris@0
|
608 writeFeatures(audioSource, plugin, featureSet,//!!! *sni);
|
Chris@0
|
609 Transform::stringToSummaryType(sni->c_str()));
|
Chris@0
|
610 }
|
Chris@0
|
611 }
|
Chris@0
|
612 }
|
Chris@0
|
613
|
Chris@0
|
614 writeSummaries(audioSource, plugin);
|
Chris@0
|
615 }
|
Chris@0
|
616
|
Chris@3
|
617 extractionProgress.done();
|
Chris@3
|
618
|
Chris@0
|
619 finish();
|
Chris@0
|
620
|
Chris@0
|
621 TempDirectory::getInstance()->cleanup();
|
Chris@0
|
622 }
|
Chris@0
|
623
|
Chris@0
|
624 void
|
Chris@0
|
625 FeatureExtractionManager::writeSummaries(QString audioSource, Plugin *plugin)
|
Chris@0
|
626 {
|
Chris@0
|
627 // caller should have ensured plugin is in m_plugins
|
Chris@0
|
628 PluginMap::iterator pi = m_plugins.find(plugin);
|
Chris@0
|
629
|
Chris@0
|
630 for (TransformWriterMap::const_iterator ti = pi->second.begin();
|
Chris@0
|
631 ti != pi->second.end(); ++ti) {
|
Chris@0
|
632
|
Chris@0
|
633 const Transform &transform = ti->first;
|
Chris@0
|
634 const vector<FeatureWriter *> &writers = ti->second;
|
Chris@0
|
635
|
Chris@0
|
636 Transform::SummaryType summaryType = transform.getSummaryType();
|
Chris@0
|
637 PluginSummarisingAdapter::SummaryType pType =
|
Chris@0
|
638 (PluginSummarisingAdapter::SummaryType)summaryType;
|
Chris@0
|
639
|
Chris@0
|
640 if (transform.getSummaryType() == Transform::NoSummary) {
|
Chris@0
|
641 continue;
|
Chris@0
|
642 }
|
Chris@0
|
643
|
Chris@0
|
644 PluginSummarisingAdapter *adapter =
|
Chris@0
|
645 dynamic_cast<PluginSummarisingAdapter *>(plugin);
|
Chris@0
|
646 if (!adapter) {
|
Chris@0
|
647 cerr << "FeatureExtractionManager::writeSummaries: INTERNAL ERROR: Summary requested for transform, but plugin is not a summarising adapter" << endl;
|
Chris@0
|
648 continue;
|
Chris@0
|
649 }
|
Chris@0
|
650
|
Chris@0
|
651 Plugin::FeatureSet featureSet = adapter->getSummaryForAllOutputs
|
Chris@0
|
652 (pType, PluginSummarisingAdapter::ContinuousTimeAverage);
|
Chris@0
|
653
|
Chris@0
|
654 // cout << "summary type " << int(pType) << " for transform:" << endl << transform.toXmlString().toStdString()<< endl << "... feature set with " << featureSet.size() << " elts" << endl;
|
Chris@0
|
655
|
Chris@0
|
656 writeFeatures(audioSource, plugin, featureSet, summaryType);
|
Chris@0
|
657 }
|
Chris@0
|
658 }
|
Chris@0
|
659
|
Chris@0
|
660 void FeatureExtractionManager::writeFeatures(QString audioSource,
|
Chris@0
|
661 Plugin *plugin,
|
Chris@0
|
662 const Plugin::FeatureSet &features,
|
Chris@0
|
663 Transform::SummaryType summaryType)
|
Chris@0
|
664 {
|
Chris@0
|
665 // caller should have ensured plugin is in m_plugins
|
Chris@0
|
666 PluginMap::iterator pi = m_plugins.find(plugin);
|
Chris@0
|
667
|
Chris@0
|
668 for (TransformWriterMap::const_iterator ti = pi->second.begin();
|
Chris@0
|
669 ti != pi->second.end(); ++ti) {
|
Chris@0
|
670
|
Chris@0
|
671 const Transform &transform = ti->first;
|
Chris@0
|
672 const vector<FeatureWriter *> &writers = ti->second;
|
Chris@0
|
673
|
Chris@0
|
674 if (transform.getSummaryType() != Transform::NoSummary &&
|
Chris@0
|
675 m_summaries.empty() &&
|
Chris@0
|
676 summaryType == Transform::NoSummary) {
|
Chris@0
|
677 continue;
|
Chris@0
|
678 }
|
Chris@0
|
679
|
Chris@0
|
680 if (transform.getSummaryType() != Transform::NoSummary &&
|
Chris@0
|
681 summaryType != Transform::NoSummary &&
|
Chris@0
|
682 transform.getSummaryType() != summaryType) {
|
Chris@0
|
683 continue;
|
Chris@0
|
684 }
|
Chris@0
|
685
|
Chris@0
|
686 string outputId = transform.getOutput().toStdString();
|
Chris@0
|
687
|
Chris@0
|
688 if (m_pluginOutputs[plugin].find(outputId) ==
|
Chris@0
|
689 m_pluginOutputs[plugin].end()) {
|
Chris@0
|
690 continue;
|
Chris@0
|
691 }
|
Chris@0
|
692
|
Chris@0
|
693 const Plugin::OutputDescriptor &desc =
|
Chris@0
|
694 m_pluginOutputs[plugin][outputId];
|
Chris@0
|
695
|
Chris@0
|
696 int outputIndex = m_pluginOutputIndices[outputId];
|
Chris@0
|
697 Plugin::FeatureSet::const_iterator fsi = features.find(outputIndex);
|
Chris@0
|
698 if (fsi == features.end()) continue;
|
Chris@0
|
699
|
Chris@0
|
700 for (int j = 0; j < (int)writers.size(); ++j) {
|
Chris@0
|
701 writers[j]->write
|
Chris@0
|
702 (audioSource, transform, desc, fsi->second,
|
Chris@0
|
703 Transform::summaryTypeToString(summaryType).toStdString());
|
Chris@0
|
704 }
|
Chris@0
|
705 }
|
Chris@0
|
706 }
|
Chris@0
|
707
|
Chris@0
|
708 void FeatureExtractionManager::finish()
|
Chris@0
|
709 {
|
Chris@0
|
710 for (PluginMap::iterator pi = m_plugins.begin();
|
Chris@0
|
711 pi != m_plugins.end(); ++pi) {
|
Chris@0
|
712
|
Chris@0
|
713 for (TransformWriterMap::iterator ti = pi->second.begin();
|
Chris@0
|
714 ti != pi->second.end(); ++ti) {
|
Chris@0
|
715
|
Chris@0
|
716 vector<FeatureWriter *> &writers = ti->second;
|
Chris@0
|
717
|
Chris@0
|
718 for (int i = 0; i < (int)writers.size(); ++i) {
|
Chris@0
|
719 writers[i]->flush();
|
Chris@0
|
720 writers[i]->finish();
|
Chris@0
|
721 }
|
Chris@0
|
722 }
|
Chris@0
|
723 }
|
Chris@0
|
724 }
|
Chris@0
|
725
|
Chris@0
|
726 void FeatureExtractionManager::print(Transform transform) const
|
Chris@0
|
727 {
|
Chris@0
|
728 QString qs;
|
Chris@0
|
729 QTextStream qts(&qs);
|
Chris@0
|
730 transform.toXml(qts);
|
Chris@0
|
731 cerr << qs.toStdString() << endl;
|
Chris@0
|
732 }
|