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@21
|
25 #include "base/Exceptions.h"
|
Chris@21
|
26
|
Chris@0
|
27 #include <iostream>
|
Chris@0
|
28
|
Chris@0
|
29 using namespace std;
|
Chris@0
|
30
|
Chris@0
|
31 using Vamp::Plugin;
|
Chris@0
|
32 using Vamp::PluginBase;
|
Chris@0
|
33 using Vamp::HostExt::PluginLoader;
|
Chris@0
|
34 using Vamp::HostExt::PluginChannelAdapter;
|
Chris@0
|
35 using Vamp::HostExt::PluginBufferingAdapter;
|
Chris@0
|
36 using Vamp::HostExt::PluginInputDomainAdapter;
|
Chris@0
|
37 using Vamp::HostExt::PluginSummarisingAdapter;
|
Chris@8
|
38 using Vamp::HostExt::PluginWrapper;
|
Chris@0
|
39
|
Chris@0
|
40 #include "data/fileio/FileSource.h"
|
Chris@0
|
41 #include "data/fileio/AudioFileReader.h"
|
Chris@0
|
42 #include "data/fileio/AudioFileReaderFactory.h"
|
Chris@0
|
43 #include "data/fileio/PlaylistFileReader.h"
|
Chris@0
|
44 #include "base/TempDirectory.h"
|
Chris@0
|
45 #include "base/ProgressPrinter.h"
|
Chris@0
|
46 #include "transform/TransformFactory.h"
|
Chris@0
|
47 #include "rdf/RDFTransformFactory.h"
|
Chris@0
|
48 #include "transform/FeatureWriter.h"
|
Chris@0
|
49
|
Chris@0
|
50 #include <QTextStream>
|
Chris@0
|
51 #include <QFile>
|
Chris@0
|
52 #include <QFileInfo>
|
Chris@0
|
53
|
Chris@0
|
54 FeatureExtractionManager::FeatureExtractionManager() :
|
Chris@0
|
55 m_summariesOnly(false),
|
Chris@0
|
56 // We can read using an arbitrary fixed block size --
|
Chris@0
|
57 // PluginBufferingAdapter handles this for us. It's likely to be
|
Chris@0
|
58 // quicker to use larger sizes than smallish ones like 1024
|
Chris@0
|
59 m_blockSize(16384),
|
Chris@0
|
60 m_defaultSampleRate(0),
|
Chris@0
|
61 m_sampleRate(0),
|
Chris@0
|
62 m_channels(1)
|
Chris@0
|
63 {
|
Chris@0
|
64 }
|
Chris@0
|
65
|
Chris@0
|
66 FeatureExtractionManager::~FeatureExtractionManager()
|
Chris@0
|
67 {
|
Chris@0
|
68 for (PluginMap::iterator pi = m_plugins.begin();
|
Chris@0
|
69 pi != m_plugins.end(); ++pi) {
|
Chris@0
|
70 delete pi->first;
|
Chris@0
|
71 }
|
Chris@0
|
72 }
|
Chris@0
|
73
|
Chris@0
|
74 void FeatureExtractionManager::setChannels(int channels)
|
Chris@0
|
75 {
|
Chris@0
|
76 m_channels = channels;
|
Chris@0
|
77 }
|
Chris@0
|
78
|
Chris@0
|
79 void FeatureExtractionManager::setDefaultSampleRate(int sampleRate)
|
Chris@0
|
80 {
|
Chris@0
|
81 m_defaultSampleRate = sampleRate;
|
Chris@0
|
82 }
|
Chris@0
|
83
|
Chris@0
|
84 static PluginSummarisingAdapter::SummaryType
|
Chris@0
|
85 getSummaryType(string name)
|
Chris@0
|
86 {
|
Chris@0
|
87 if (name == "min") return PluginSummarisingAdapter::Minimum;
|
Chris@0
|
88 if (name == "max") return PluginSummarisingAdapter::Maximum;
|
Chris@0
|
89 if (name == "mean") return PluginSummarisingAdapter::Mean;
|
Chris@0
|
90 if (name == "median") return PluginSummarisingAdapter::Median;
|
Chris@0
|
91 if (name == "mode") return PluginSummarisingAdapter::Mode;
|
Chris@0
|
92 if (name == "sum") return PluginSummarisingAdapter::Sum;
|
Chris@0
|
93 if (name == "variance") return PluginSummarisingAdapter::Variance;
|
Chris@0
|
94 if (name == "sd") return PluginSummarisingAdapter::StandardDeviation;
|
Chris@0
|
95 if (name == "count") return PluginSummarisingAdapter::Count;
|
Chris@0
|
96 return PluginSummarisingAdapter::UnknownSummaryType;
|
Chris@0
|
97 }
|
Chris@0
|
98
|
Chris@0
|
99 bool FeatureExtractionManager::setSummaryTypes(const set<string> &names,
|
Chris@0
|
100 bool summariesOnly,
|
Chris@0
|
101 const PluginSummarisingAdapter::SegmentBoundaries &boundaries)
|
Chris@0
|
102 {
|
Chris@0
|
103 for (SummaryNameSet::const_iterator i = names.begin();
|
Chris@0
|
104 i != names.end(); ++i) {
|
Chris@0
|
105 if (getSummaryType(*i) == PluginSummarisingAdapter::UnknownSummaryType) {
|
Chris@0
|
106 cerr << "ERROR: Unknown summary type \"" << *i << "\"" << endl;
|
Chris@0
|
107 return false;
|
Chris@0
|
108 }
|
Chris@0
|
109 }
|
Chris@0
|
110 m_summaries = names;
|
Chris@0
|
111 m_summariesOnly = summariesOnly;
|
Chris@0
|
112 m_boundaries = boundaries;
|
Chris@0
|
113 return true;
|
Chris@0
|
114 }
|
Chris@0
|
115
|
Chris@0
|
116 bool FeatureExtractionManager::addFeatureExtractor
|
Chris@0
|
117 (Transform transform, const vector<FeatureWriter*> &writers)
|
Chris@0
|
118 {
|
Chris@0
|
119 //!!! exceptions rather than return values?
|
Chris@0
|
120
|
Chris@0
|
121 if (transform.getSampleRate() == 0) {
|
Chris@0
|
122 if (m_sampleRate == 0) {
|
Chris@0
|
123 cerr << "NOTE: Transform does not specify a sample rate, using default rate of " << m_defaultSampleRate << endl;
|
Chris@0
|
124 transform.setSampleRate(m_defaultSampleRate);
|
Chris@0
|
125 m_sampleRate = m_defaultSampleRate;
|
Chris@0
|
126 } else {
|
Chris@0
|
127 cerr << "NOTE: Transform does not specify a sample rate, using previous transform's rate of " << m_sampleRate << endl;
|
Chris@0
|
128 transform.setSampleRate(m_sampleRate);
|
Chris@0
|
129 }
|
Chris@0
|
130 }
|
Chris@0
|
131
|
Chris@0
|
132 if (m_sampleRate == 0) {
|
Chris@0
|
133 m_sampleRate = transform.getSampleRate();
|
Chris@0
|
134 }
|
Chris@0
|
135
|
Chris@0
|
136 if (transform.getSampleRate() != m_sampleRate) {
|
Chris@0
|
137 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
|
138 cerr << "WARNING: Using previous rate of " << m_sampleRate << " for this transform as well" << endl;
|
Chris@0
|
139 transform.setSampleRate(m_sampleRate);
|
Chris@0
|
140 }
|
Chris@0
|
141
|
Chris@0
|
142 Plugin *plugin = 0;
|
Chris@0
|
143
|
Chris@0
|
144 // Remember what the original transform looked like, and index
|
Chris@0
|
145 // based on this -- because we may be about to fill in the zeros
|
Chris@0
|
146 // for step and block size, but we want any further copies with
|
Chris@0
|
147 // the same zeros to match this one
|
Chris@0
|
148 Transform originalTransform = transform;
|
Chris@0
|
149
|
Chris@0
|
150 if (m_transformPluginMap.find(transform) == m_transformPluginMap.end()) {
|
Chris@0
|
151
|
Chris@0
|
152 // Test whether we already have a transform that is identical
|
Chris@0
|
153 // to this, except for the output requested and/or the summary
|
Chris@0
|
154 // type -- if so, they should share plugin instances (a vital
|
Chris@0
|
155 // optimisation)
|
Chris@0
|
156
|
Chris@0
|
157 for (TransformPluginMap::iterator i = m_transformPluginMap.begin();
|
Chris@0
|
158 i != m_transformPluginMap.end(); ++i) {
|
Chris@0
|
159 Transform test = i->first;
|
Chris@0
|
160 test.setOutput(transform.getOutput());
|
Chris@0
|
161 test.setSummaryType(transform.getSummaryType());
|
Chris@0
|
162 if (transform == test) {
|
Chris@0
|
163 cerr << "NOTE: Already have transform identical to this one (for \""
|
Chris@0
|
164 << transform.getIdentifier().toStdString()
|
Chris@0
|
165 << "\") in every detail except output identifier and/or "
|
Chris@0
|
166 << "summary type; sharing its plugin instance" << endl;
|
Chris@0
|
167 plugin = i->second;
|
Chris@0
|
168 if (transform.getSummaryType() != Transform::NoSummary &&
|
Chris@0
|
169 !dynamic_cast<PluginSummarisingAdapter *>(plugin)) {
|
Chris@0
|
170 plugin = new PluginSummarisingAdapter(plugin);
|
Chris@0
|
171 i->second = plugin;
|
Chris@0
|
172 }
|
Chris@0
|
173 break;
|
Chris@0
|
174 }
|
Chris@0
|
175 }
|
Chris@0
|
176
|
Chris@0
|
177 if (!plugin) {
|
Chris@0
|
178
|
Chris@0
|
179 TransformFactory *tf = TransformFactory::getInstance();
|
Chris@0
|
180
|
Chris@0
|
181 PluginBase *pb = tf->instantiatePluginFor(transform);
|
Chris@0
|
182 plugin = tf->downcastVampPlugin(pb);
|
Chris@0
|
183 if (!plugin) {
|
Chris@0
|
184 //!!! todo: handle non-Vamp plugins too, or make the main --list
|
Chris@0
|
185 // option print out only Vamp transforms
|
Chris@0
|
186 cerr << "ERROR: Failed to load plugin for transform \""
|
Chris@0
|
187 << transform.getIdentifier().toStdString() << "\"" << endl;
|
Chris@0
|
188 delete pb;
|
Chris@0
|
189 return false;
|
Chris@0
|
190 }
|
Chris@0
|
191
|
Chris@0
|
192 // We will provide the plugin with arbitrary step and
|
Chris@0
|
193 // block sizes (so that we can use the same read/write
|
Chris@0
|
194 // block size for all transforms), and to that end we use
|
Chris@0
|
195 // a PluginBufferingAdapter. However, we need to know the
|
Chris@0
|
196 // underlying step size so that we can provide the right
|
Chris@0
|
197 // context for dense outputs. (Although, don't forget
|
Chris@0
|
198 // that the PluginBufferingAdapter rewrites
|
Chris@0
|
199 // OneSamplePerStep outputs so as to use FixedSampleRate
|
Chris@0
|
200 // -- so it supplies the sample rate in the output
|
Chris@0
|
201 // feature. I'm not sure whether we can easily use that.)
|
Chris@0
|
202
|
Chris@0
|
203 size_t pluginStepSize = plugin->getPreferredStepSize();
|
Chris@0
|
204 size_t pluginBlockSize = plugin->getPreferredBlockSize();
|
Chris@0
|
205
|
Chris@25
|
206 PluginInputDomainAdapter *pida = 0;
|
Chris@25
|
207
|
Chris@0
|
208 // adapt the plugin for buffering, channels, etc.
|
Chris@0
|
209 if (plugin->getInputDomain() == Plugin::FrequencyDomain) {
|
Chris@25
|
210 pida = new PluginInputDomainAdapter(plugin);
|
Chris@25
|
211 plugin = pida;
|
Chris@0
|
212 }
|
Chris@0
|
213
|
Chris@0
|
214 PluginBufferingAdapter *pba = new PluginBufferingAdapter(plugin);
|
Chris@0
|
215 plugin = pba;
|
Chris@0
|
216
|
Chris@0
|
217 if (transform.getStepSize() != 0) {
|
Chris@0
|
218 pba->setPluginStepSize(transform.getStepSize());
|
Chris@0
|
219 } else {
|
Chris@0
|
220 transform.setStepSize(pluginStepSize);
|
Chris@0
|
221 }
|
Chris@0
|
222
|
Chris@0
|
223 if (transform.getBlockSize() != 0) {
|
Chris@0
|
224 pba->setPluginBlockSize(transform.getBlockSize());
|
Chris@0
|
225 } else {
|
Chris@0
|
226 transform.setBlockSize(pluginBlockSize);
|
Chris@0
|
227 }
|
Chris@0
|
228
|
Chris@0
|
229 plugin = new PluginChannelAdapter(plugin);
|
Chris@0
|
230
|
Chris@0
|
231 if (!m_summaries.empty() ||
|
Chris@0
|
232 transform.getSummaryType() != Transform::NoSummary) {
|
Chris@0
|
233 PluginSummarisingAdapter *adapter =
|
Chris@0
|
234 new PluginSummarisingAdapter(plugin);
|
Chris@0
|
235 adapter->setSummarySegmentBoundaries(m_boundaries);
|
Chris@0
|
236 plugin = adapter;
|
Chris@0
|
237 }
|
Chris@0
|
238
|
Chris@0
|
239 if (!plugin->initialise(m_channels, m_blockSize, m_blockSize)) {
|
Chris@0
|
240 cerr << "ERROR: Plugin initialise (channels = " << m_channels << ", stepSize = " << m_blockSize << ", blockSize = " << m_blockSize << ") failed." << endl;
|
Chris@0
|
241 delete plugin;
|
Chris@0
|
242 return false;
|
Chris@0
|
243 }
|
Chris@0
|
244
|
Chris@0
|
245 // cerr << "Initialised plugin" << endl;
|
Chris@0
|
246
|
Chris@0
|
247 size_t actualStepSize = 0;
|
Chris@0
|
248 size_t actualBlockSize = 0;
|
Chris@0
|
249 pba->getActualStepAndBlockSizes(actualStepSize, actualBlockSize);
|
Chris@0
|
250 transform.setStepSize(actualStepSize);
|
Chris@0
|
251 transform.setBlockSize(actualBlockSize);
|
Chris@0
|
252
|
Chris@0
|
253 Plugin::OutputList outputs = plugin->getOutputDescriptors();
|
Chris@0
|
254 for (int i = 0; i < (int)outputs.size(); ++i) {
|
Chris@0
|
255
|
Chris@0
|
256 // cerr << "Newly initialised plugin output " << i << " has bin count " << outputs[i].binCount << endl;
|
Chris@0
|
257
|
Chris@0
|
258 m_pluginOutputs[plugin][outputs[i].identifier] = outputs[i];
|
Chris@0
|
259 m_pluginOutputIndices[outputs[i].identifier] = i;
|
Chris@0
|
260 }
|
Chris@0
|
261
|
Chris@10
|
262 cerr << "NOTE: Loaded and initialised plugin for transform \""
|
Chris@25
|
263 << transform.getIdentifier().toStdString()
|
Chris@25
|
264 << "\" with plugin step size " << actualStepSize
|
Chris@25
|
265 << " and block size " << actualBlockSize
|
Chris@25
|
266 << " (adapter step and block size " << m_blockSize << ")"
|
Chris@25
|
267 << endl;
|
Chris@25
|
268
|
Chris@25
|
269 if (pida) {
|
Chris@25
|
270 cerr << "NOTE: PluginInputDomainAdapter timestamp adjustment is "
|
Chris@25
|
271
|
Chris@25
|
272 << pida->getTimestampAdjustment() << endl;
|
Chris@25
|
273 }
|
Chris@8
|
274
|
Chris@8
|
275 } else {
|
Chris@8
|
276
|
Chris@8
|
277 if (transform.getStepSize() == 0 || transform.getBlockSize() == 0) {
|
Chris@8
|
278
|
Chris@8
|
279 PluginWrapper *pw = dynamic_cast<PluginWrapper *>(plugin);
|
Chris@8
|
280 if (pw) {
|
Chris@8
|
281 PluginBufferingAdapter *pba =
|
Chris@8
|
282 pw->getWrapper<PluginBufferingAdapter>();
|
Chris@8
|
283 if (pba) {
|
Chris@8
|
284 size_t actualStepSize = 0;
|
Chris@8
|
285 size_t actualBlockSize = 0;
|
Chris@8
|
286 pba->getActualStepAndBlockSizes(actualStepSize,
|
Chris@8
|
287 actualBlockSize);
|
Chris@8
|
288 if (transform.getStepSize() == 0) {
|
Chris@8
|
289 transform.setStepSize(actualStepSize);
|
Chris@8
|
290 }
|
Chris@8
|
291 if (transform.getBlockSize() == 0) {
|
Chris@8
|
292 transform.setBlockSize(actualBlockSize);
|
Chris@8
|
293 }
|
Chris@8
|
294 }
|
Chris@8
|
295 }
|
Chris@8
|
296 }
|
Chris@0
|
297 }
|
Chris@0
|
298
|
Chris@0
|
299 if (transform.getOutput() == "") {
|
Chris@0
|
300 transform.setOutput
|
Chris@0
|
301 (plugin->getOutputDescriptors()[0].identifier.c_str());
|
Chris@0
|
302 }
|
Chris@0
|
303
|
Chris@0
|
304 m_transformPluginMap[transform] = plugin;
|
Chris@0
|
305
|
Chris@0
|
306 if (!(originalTransform == transform)) {
|
Chris@0
|
307 m_transformPluginMap[originalTransform] = plugin;
|
Chris@0
|
308 }
|
Chris@0
|
309
|
Chris@0
|
310 } else {
|
Chris@0
|
311
|
Chris@0
|
312 plugin = m_transformPluginMap[transform];
|
Chris@0
|
313 }
|
Chris@0
|
314
|
Chris@0
|
315 m_plugins[plugin][transform] = writers;
|
Chris@0
|
316
|
Chris@0
|
317 return true;
|
Chris@0
|
318 }
|
Chris@0
|
319
|
Chris@0
|
320 bool FeatureExtractionManager::addDefaultFeatureExtractor
|
Chris@0
|
321 (TransformId transformId, const vector<FeatureWriter*> &writers)
|
Chris@0
|
322 {
|
Chris@0
|
323 TransformFactory *tf = TransformFactory::getInstance();
|
Chris@0
|
324
|
Chris@0
|
325 if (m_sampleRate == 0) {
|
Chris@0
|
326 if (m_defaultSampleRate == 0) {
|
Chris@0
|
327 cerr << "ERROR: Default transform requested, but no default sample rate available" << endl;
|
Chris@0
|
328 return false;
|
Chris@0
|
329 } else {
|
Chris@0
|
330 cerr << "NOTE: Using default sample rate of " << m_defaultSampleRate << " for default transform" << endl;
|
Chris@0
|
331 m_sampleRate = m_defaultSampleRate;
|
Chris@0
|
332 }
|
Chris@0
|
333 }
|
Chris@0
|
334
|
Chris@0
|
335 Transform transform = tf->getDefaultTransformFor(transformId, m_sampleRate);
|
Chris@0
|
336
|
Chris@0
|
337 return addFeatureExtractor(transform, writers);
|
Chris@0
|
338 }
|
Chris@0
|
339
|
Chris@0
|
340 bool FeatureExtractionManager::addFeatureExtractorFromFile
|
Chris@0
|
341 (QString transformXmlFile, const vector<FeatureWriter*> &writers)
|
Chris@0
|
342 {
|
Chris@0
|
343 RDFTransformFactory factory
|
Chris@0
|
344 (QUrl::fromLocalFile(QFileInfo(transformXmlFile).absoluteFilePath())
|
Chris@0
|
345 .toString());
|
Chris@0
|
346 ProgressPrinter printer("Parsing transforms RDF file");
|
Chris@0
|
347 std::vector<Transform> transforms = factory.getTransforms(&printer);
|
Chris@0
|
348 if (!factory.isOK()) {
|
Chris@0
|
349 cerr << "WARNING: FeatureExtractionManager::addFeatureExtractorFromFile: Failed to parse transforms file: " << factory.getErrorString().toStdString() << endl;
|
Chris@0
|
350 if (factory.isRDF()) {
|
Chris@0
|
351 return false; // no point trying it as XML
|
Chris@0
|
352 }
|
Chris@0
|
353 }
|
Chris@0
|
354 if (!transforms.empty()) {
|
Chris@0
|
355 bool success = true;
|
Chris@0
|
356 for (int i = 0; i < (int)transforms.size(); ++i) {
|
Chris@0
|
357 if (!addFeatureExtractor(transforms[i], writers)) {
|
Chris@0
|
358 success = false;
|
Chris@0
|
359 }
|
Chris@0
|
360 }
|
Chris@0
|
361 return success;
|
Chris@0
|
362 }
|
Chris@0
|
363
|
Chris@0
|
364 QFile file(transformXmlFile);
|
Chris@0
|
365 if (!file.open(QIODevice::ReadOnly | QIODevice::Text)) {
|
Chris@0
|
366 cerr << "ERROR: Failed to open transform XML file \""
|
Chris@0
|
367 << transformXmlFile.toStdString() << "\" for reading" << endl;
|
Chris@0
|
368 return false;
|
Chris@0
|
369 }
|
Chris@0
|
370
|
Chris@0
|
371 QTextStream *qts = new QTextStream(&file);
|
Chris@0
|
372 QString qs = qts->readAll();
|
Chris@0
|
373 delete qts;
|
Chris@0
|
374 file.close();
|
Chris@0
|
375
|
Chris@0
|
376 Transform transform(qs);
|
Chris@0
|
377
|
Chris@0
|
378 return addFeatureExtractor(transform, writers);
|
Chris@0
|
379 }
|
Chris@0
|
380
|
Chris@0
|
381 void FeatureExtractionManager::extractFeatures(QString audioSource)
|
Chris@0
|
382 {
|
Chris@0
|
383 if (m_plugins.empty()) return;
|
Chris@0
|
384
|
Chris@11
|
385 ProgressPrinter retrievalProgress("Retrieving audio data...");
|
Chris@0
|
386
|
Chris@11
|
387 FileSource source(audioSource, &retrievalProgress);
|
Chris@0
|
388 if (!source.isAvailable()) {
|
Chris@0
|
389 cerr << "ERROR: File or URL \"" << audioSource.toStdString()
|
Chris@0
|
390 << "\" could not be located" << endl;
|
Chris@21
|
391 throw FileNotFound(audioSource);
|
Chris@0
|
392 }
|
Chris@0
|
393
|
Chris@0
|
394 source.waitForData();
|
Chris@0
|
395
|
Chris@0
|
396 if (QFileInfo(audioSource).suffix().toLower() == "m3u") {
|
Chris@0
|
397 PlaylistFileReader reader(source);
|
Chris@0
|
398 if (reader.isOK()) {
|
Chris@0
|
399 vector<QString> files = reader.load();
|
Chris@0
|
400 for (int i = 0; i < (int)files.size(); ++i) {
|
Chris@0
|
401 extractFeatures(files[i]);
|
Chris@0
|
402 }
|
Chris@0
|
403 return;
|
Chris@0
|
404 } else {
|
Chris@0
|
405 cerr << "ERROR: Playlist \"" << audioSource.toStdString()
|
Chris@0
|
406 << "\" could not be opened" << endl;
|
Chris@21
|
407 throw FileNotFound(audioSource);
|
Chris@0
|
408 }
|
Chris@0
|
409 }
|
Chris@0
|
410
|
Chris@0
|
411 if (m_sampleRate == 0) {
|
Chris@0
|
412 cerr << "ERROR: Internal error in FeatureExtractionManager::extractFeatures: Plugin list is non-empty, but no sample rate set" << endl;
|
Chris@0
|
413 exit(1);
|
Chris@0
|
414 }
|
Chris@0
|
415
|
Chris@0
|
416 AudioFileReader *reader =
|
Chris@11
|
417 AudioFileReaderFactory::createReader(source, m_sampleRate, &retrievalProgress);
|
Chris@0
|
418
|
Chris@0
|
419 if (!reader) {
|
Chris@21
|
420 throw FailedToOpenFile(audioSource);
|
Chris@0
|
421 }
|
Chris@0
|
422
|
Chris@0
|
423 size_t channels = reader->getChannelCount();
|
Chris@11
|
424
|
Chris@11
|
425 retrievalProgress.done();
|
Chris@11
|
426
|
Chris@0
|
427 cerr << "Opened " << channels << "-channel file or URL \"" << audioSource.toStdString() << "\"" << endl;
|
Chris@0
|
428
|
Chris@0
|
429 // reject file if it has too few channels, plugin will handle if it has too many
|
Chris@0
|
430 if ((int)channels < m_channels) {
|
Chris@21
|
431 throw FileOperationFailed
|
Chris@21
|
432 (audioSource,
|
Chris@21
|
433 QString("read sufficient channels (found %1, require %2)")
|
Chris@21
|
434 .arg(channels).arg(m_channels));
|
Chris@0
|
435 }
|
Chris@0
|
436
|
Chris@0
|
437 // allocate audio buffers
|
Chris@0
|
438 float **data = new float *[m_channels];
|
Chris@0
|
439 for (int c = 0; c < m_channels; ++c) {
|
Chris@0
|
440 data[c] = new float[m_blockSize];
|
Chris@0
|
441 }
|
Chris@0
|
442
|
Chris@0
|
443 size_t frameCount = reader->getFrameCount();
|
Chris@0
|
444
|
Chris@0
|
445 // cerr << "file has " << frameCount << " frames" << endl;
|
Chris@0
|
446
|
Chris@0
|
447 for (PluginMap::iterator pi = m_plugins.begin();
|
Chris@0
|
448 pi != m_plugins.end(); ++pi) {
|
Chris@0
|
449
|
Chris@0
|
450 Plugin *plugin = pi->first;
|
Chris@0
|
451
|
Chris@0
|
452 // std::cerr << "Calling reset on " << plugin << std::endl;
|
Chris@0
|
453 plugin->reset();
|
Chris@0
|
454
|
Chris@0
|
455 for (TransformWriterMap::iterator ti = pi->second.begin();
|
Chris@0
|
456 ti != pi->second.end(); ++ti) {
|
Chris@0
|
457
|
Chris@0
|
458 const Transform &transform = ti->first;
|
Chris@0
|
459
|
Chris@0
|
460 //!!! we may want to set the start and duration times for extraction
|
Chris@0
|
461 // in the transform record (defaults of zero indicate extraction
|
Chris@0
|
462 // from the whole file)
|
Chris@0
|
463 // transform.setStartTime(RealTime::zeroTime);
|
Chris@0
|
464 // transform.setDuration
|
Chris@0
|
465 // (RealTime::frame2RealTime(reader->getFrameCount(), m_sampleRate));
|
Chris@0
|
466
|
Chris@0
|
467 string outputId = transform.getOutput().toStdString();
|
Chris@0
|
468 if (m_pluginOutputs[plugin].find(outputId) ==
|
Chris@0
|
469 m_pluginOutputs[plugin].end()) {
|
Chris@0
|
470 //!!! throw?
|
Chris@0
|
471 cerr << "WARNING: Nonexistent plugin output \"" << outputId << "\" requested for transform \""
|
Chris@0
|
472 << transform.getIdentifier().toStdString() << "\", ignoring this transform"
|
Chris@0
|
473 << endl;
|
Chris@0
|
474 /*
|
Chris@0
|
475 cerr << "Known outputs for all plugins are as follows:" << endl;
|
Chris@0
|
476 for (PluginOutputMap::const_iterator k = m_pluginOutputs.begin();
|
Chris@0
|
477 k != m_pluginOutputs.end(); ++k) {
|
Chris@0
|
478 cerr << "Plugin " << k->first << ": ";
|
Chris@0
|
479 if (k->second.empty()) {
|
Chris@0
|
480 cerr << "(none)";
|
Chris@0
|
481 }
|
Chris@0
|
482 for (OutputMap::const_iterator i = k->second.begin();
|
Chris@0
|
483 i != k->second.end(); ++i) {
|
Chris@0
|
484 cerr << "\"" << i->first << "\" ";
|
Chris@0
|
485 }
|
Chris@0
|
486 cerr << endl;
|
Chris@0
|
487 }
|
Chris@0
|
488 */
|
Chris@0
|
489 }
|
Chris@0
|
490 }
|
Chris@0
|
491 }
|
Chris@0
|
492
|
Chris@0
|
493 long startFrame = 0;
|
Chris@0
|
494 long endFrame = frameCount;
|
Chris@0
|
495
|
Chris@0
|
496 /*!!! No -- there is no single transform to pull this stuff from --
|
Chris@0
|
497 * the transforms may have various start and end times, need to be far
|
Chris@0
|
498 * cleverer about this if we're going to support them
|
Chris@0
|
499
|
Chris@0
|
500 RealTime trStartRT = transform.getStartTime();
|
Chris@0
|
501 RealTime trDurationRT = transform.getDuration();
|
Chris@0
|
502
|
Chris@0
|
503 long trStart = RealTime::realTime2Frame(trStartRT, m_sampleRate);
|
Chris@0
|
504 long trDuration = RealTime::realTime2Frame(trDurationRT, m_sampleRate);
|
Chris@0
|
505
|
Chris@0
|
506 if (trStart == 0 || trStart < startFrame) {
|
Chris@0
|
507 trStart = startFrame;
|
Chris@0
|
508 }
|
Chris@0
|
509
|
Chris@0
|
510 if (trDuration == 0) {
|
Chris@0
|
511 trDuration = endFrame - trStart;
|
Chris@0
|
512 }
|
Chris@0
|
513 if (trStart + trDuration > endFrame) {
|
Chris@0
|
514 trDuration = endFrame - trStart;
|
Chris@0
|
515 }
|
Chris@0
|
516
|
Chris@0
|
517 startFrame = trStart;
|
Chris@0
|
518 endFrame = trStart + trDuration;
|
Chris@0
|
519 */
|
Chris@0
|
520
|
Chris@0
|
521 for (PluginMap::iterator pi = m_plugins.begin();
|
Chris@0
|
522 pi != m_plugins.end(); ++pi) {
|
Chris@0
|
523
|
Chris@0
|
524 for (TransformWriterMap::const_iterator ti = pi->second.begin();
|
Chris@0
|
525 ti != pi->second.end(); ++ti) {
|
Chris@0
|
526
|
Chris@0
|
527 const vector<FeatureWriter *> &writers = ti->second;
|
Chris@0
|
528
|
Chris@0
|
529 for (int j = 0; j < (int)writers.size(); ++j) {
|
Chris@0
|
530 FeatureWriter::TrackMetadata m;
|
Chris@0
|
531 m.title = reader->getTitle();
|
Chris@0
|
532 m.maker = reader->getMaker();
|
Chris@19
|
533 if (m.title != "" && m.maker != "") {
|
Chris@19
|
534 writers[j]->setTrackMetadata(audioSource, m);
|
Chris@19
|
535 }
|
Chris@0
|
536 }
|
Chris@0
|
537 }
|
Chris@0
|
538 }
|
Chris@0
|
539
|
Chris@0
|
540 ProgressPrinter extractionProgress("Extracting and writing features...");
|
Chris@0
|
541 int progress = 0;
|
Chris@0
|
542
|
Chris@0
|
543 for (long i = startFrame; i < endFrame; i += m_blockSize) {
|
Chris@0
|
544
|
Chris@0
|
545 //!!! inefficient, although much of the inefficiency may be
|
Chris@0
|
546 // susceptible to optimisation
|
Chris@0
|
547
|
Chris@0
|
548 SampleBlock frames;
|
Chris@0
|
549 reader->getInterleavedFrames(i, m_blockSize, frames);
|
Chris@0
|
550
|
Chris@0
|
551 // We have to do our own channel handling here; we can't just
|
Chris@0
|
552 // leave it to the plugin adapter because the same plugin
|
Chris@0
|
553 // adapter may have to serve for input files with various
|
Chris@0
|
554 // numbers of channels (so the adapter is simply configured
|
Chris@0
|
555 // with a fixed channel count, generally 1).
|
Chris@0
|
556
|
Chris@0
|
557 int rc = reader->getChannelCount();
|
Chris@0
|
558
|
Chris@0
|
559 for (int j = 0; j < m_blockSize; ++j) {
|
Chris@0
|
560 for (int c = 0; c < m_channels; ++c) {
|
Chris@0
|
561 int index;
|
Chris@0
|
562 if (c < rc) {
|
Chris@0
|
563 index = j * rc + c;
|
Chris@0
|
564 data[c][j] = 0.f;
|
Chris@0
|
565 } else {
|
Chris@0
|
566 index = j * rc + (c % rc);
|
Chris@0
|
567 }
|
Chris@0
|
568 if (index < (int)frames.size()) {
|
Chris@0
|
569 data[c][j] += frames[index];
|
Chris@0
|
570 }
|
Chris@0
|
571 }
|
Chris@0
|
572 }
|
Chris@0
|
573
|
Chris@0
|
574 Vamp::RealTime timestamp = Vamp::RealTime::frame2RealTime
|
Chris@0
|
575 (i, m_sampleRate);
|
Chris@0
|
576
|
Chris@0
|
577 for (PluginMap::iterator pi = m_plugins.begin();
|
Chris@0
|
578 pi != m_plugins.end(); ++pi) {
|
Chris@0
|
579
|
Chris@0
|
580 Plugin *plugin = pi->first;
|
Chris@0
|
581 Plugin::FeatureSet featureSet = plugin->process(data, timestamp);
|
Chris@0
|
582
|
Chris@0
|
583 if (!m_summariesOnly) {
|
Chris@0
|
584 writeFeatures(audioSource, plugin, featureSet);
|
Chris@0
|
585 }
|
Chris@0
|
586 }
|
Chris@0
|
587
|
Chris@0
|
588 int pp = progress;
|
Chris@6
|
589 progress = int(((i - startFrame) * 100.0) / (endFrame - startFrame) + 0.1);
|
Chris@0
|
590 if (progress > pp) extractionProgress.setProgress(progress);
|
Chris@0
|
591 }
|
Chris@10
|
592
|
Chris@22
|
593 // std::cerr << "FeatureExtractionManager: deleting audio file reader" << std::endl;
|
Chris@12
|
594
|
Chris@10
|
595 delete reader;
|
Chris@0
|
596
|
Chris@0
|
597 for (PluginMap::iterator pi = m_plugins.begin();
|
Chris@0
|
598 pi != m_plugins.end(); ++pi) {
|
Chris@0
|
599
|
Chris@0
|
600 Plugin *plugin = pi->first;
|
Chris@0
|
601 Plugin::FeatureSet featureSet = plugin->getRemainingFeatures();
|
Chris@0
|
602
|
Chris@0
|
603 if (!m_summariesOnly) {
|
Chris@0
|
604 writeFeatures(audioSource, plugin, featureSet);
|
Chris@0
|
605 }
|
Chris@0
|
606
|
Chris@0
|
607 if (!m_summaries.empty()) {
|
Chris@0
|
608 PluginSummarisingAdapter *adapter =
|
Chris@0
|
609 dynamic_cast<PluginSummarisingAdapter *>(plugin);
|
Chris@0
|
610 if (!adapter) {
|
Chris@0
|
611 cerr << "WARNING: Summaries requested, but plugin is not a summarising adapter" << endl;
|
Chris@0
|
612 } else {
|
Chris@0
|
613 for (SummaryNameSet::const_iterator sni = m_summaries.begin();
|
Chris@0
|
614 sni != m_summaries.end(); ++sni) {
|
Chris@0
|
615 featureSet.clear();
|
Chris@0
|
616 //!!! problem here -- we are requesting summaries
|
Chris@0
|
617 //!!! for all outputs, but they in principle have
|
Chris@0
|
618 //!!! different averaging requirements depending
|
Chris@0
|
619 //!!! on whether their features have duration or
|
Chris@0
|
620 //!!! not
|
Chris@0
|
621 featureSet = adapter->getSummaryForAllOutputs
|
Chris@0
|
622 (getSummaryType(*sni),
|
Chris@0
|
623 PluginSummarisingAdapter::ContinuousTimeAverage);
|
Chris@0
|
624 writeFeatures(audioSource, plugin, featureSet,//!!! *sni);
|
Chris@0
|
625 Transform::stringToSummaryType(sni->c_str()));
|
Chris@0
|
626 }
|
Chris@0
|
627 }
|
Chris@0
|
628 }
|
Chris@0
|
629
|
Chris@0
|
630 writeSummaries(audioSource, plugin);
|
Chris@0
|
631 }
|
Chris@0
|
632
|
Chris@3
|
633 extractionProgress.done();
|
Chris@3
|
634
|
Chris@0
|
635 finish();
|
Chris@0
|
636
|
Chris@0
|
637 TempDirectory::getInstance()->cleanup();
|
Chris@0
|
638 }
|
Chris@0
|
639
|
Chris@0
|
640 void
|
Chris@0
|
641 FeatureExtractionManager::writeSummaries(QString audioSource, Plugin *plugin)
|
Chris@0
|
642 {
|
Chris@0
|
643 // caller should have ensured plugin is in m_plugins
|
Chris@0
|
644 PluginMap::iterator pi = m_plugins.find(plugin);
|
Chris@0
|
645
|
Chris@0
|
646 for (TransformWriterMap::const_iterator ti = pi->second.begin();
|
Chris@0
|
647 ti != pi->second.end(); ++ti) {
|
Chris@0
|
648
|
Chris@0
|
649 const Transform &transform = ti->first;
|
Chris@0
|
650 const vector<FeatureWriter *> &writers = ti->second;
|
Chris@0
|
651
|
Chris@0
|
652 Transform::SummaryType summaryType = transform.getSummaryType();
|
Chris@0
|
653 PluginSummarisingAdapter::SummaryType pType =
|
Chris@0
|
654 (PluginSummarisingAdapter::SummaryType)summaryType;
|
Chris@0
|
655
|
Chris@0
|
656 if (transform.getSummaryType() == Transform::NoSummary) {
|
Chris@0
|
657 continue;
|
Chris@0
|
658 }
|
Chris@0
|
659
|
Chris@0
|
660 PluginSummarisingAdapter *adapter =
|
Chris@0
|
661 dynamic_cast<PluginSummarisingAdapter *>(plugin);
|
Chris@0
|
662 if (!adapter) {
|
Chris@0
|
663 cerr << "FeatureExtractionManager::writeSummaries: INTERNAL ERROR: Summary requested for transform, but plugin is not a summarising adapter" << endl;
|
Chris@0
|
664 continue;
|
Chris@0
|
665 }
|
Chris@0
|
666
|
Chris@0
|
667 Plugin::FeatureSet featureSet = adapter->getSummaryForAllOutputs
|
Chris@0
|
668 (pType, PluginSummarisingAdapter::ContinuousTimeAverage);
|
Chris@0
|
669
|
Chris@0
|
670 // cout << "summary type " << int(pType) << " for transform:" << endl << transform.toXmlString().toStdString()<< endl << "... feature set with " << featureSet.size() << " elts" << endl;
|
Chris@0
|
671
|
Chris@0
|
672 writeFeatures(audioSource, plugin, featureSet, summaryType);
|
Chris@0
|
673 }
|
Chris@0
|
674 }
|
Chris@0
|
675
|
Chris@0
|
676 void FeatureExtractionManager::writeFeatures(QString audioSource,
|
Chris@0
|
677 Plugin *plugin,
|
Chris@0
|
678 const Plugin::FeatureSet &features,
|
Chris@0
|
679 Transform::SummaryType summaryType)
|
Chris@0
|
680 {
|
Chris@0
|
681 // caller should have ensured plugin is in m_plugins
|
Chris@0
|
682 PluginMap::iterator pi = m_plugins.find(plugin);
|
Chris@0
|
683
|
Chris@0
|
684 for (TransformWriterMap::const_iterator ti = pi->second.begin();
|
Chris@0
|
685 ti != pi->second.end(); ++ti) {
|
Chris@0
|
686
|
Chris@0
|
687 const Transform &transform = ti->first;
|
Chris@0
|
688 const vector<FeatureWriter *> &writers = ti->second;
|
Chris@0
|
689
|
Chris@0
|
690 if (transform.getSummaryType() != Transform::NoSummary &&
|
Chris@0
|
691 m_summaries.empty() &&
|
Chris@0
|
692 summaryType == Transform::NoSummary) {
|
Chris@0
|
693 continue;
|
Chris@0
|
694 }
|
Chris@0
|
695
|
Chris@0
|
696 if (transform.getSummaryType() != Transform::NoSummary &&
|
Chris@0
|
697 summaryType != Transform::NoSummary &&
|
Chris@0
|
698 transform.getSummaryType() != summaryType) {
|
Chris@0
|
699 continue;
|
Chris@0
|
700 }
|
Chris@0
|
701
|
Chris@0
|
702 string outputId = transform.getOutput().toStdString();
|
Chris@0
|
703
|
Chris@0
|
704 if (m_pluginOutputs[plugin].find(outputId) ==
|
Chris@0
|
705 m_pluginOutputs[plugin].end()) {
|
Chris@0
|
706 continue;
|
Chris@0
|
707 }
|
Chris@0
|
708
|
Chris@0
|
709 const Plugin::OutputDescriptor &desc =
|
Chris@0
|
710 m_pluginOutputs[plugin][outputId];
|
Chris@0
|
711
|
Chris@0
|
712 int outputIndex = m_pluginOutputIndices[outputId];
|
Chris@0
|
713 Plugin::FeatureSet::const_iterator fsi = features.find(outputIndex);
|
Chris@0
|
714 if (fsi == features.end()) continue;
|
Chris@0
|
715
|
Chris@0
|
716 for (int j = 0; j < (int)writers.size(); ++j) {
|
Chris@0
|
717 writers[j]->write
|
Chris@0
|
718 (audioSource, transform, desc, fsi->second,
|
Chris@0
|
719 Transform::summaryTypeToString(summaryType).toStdString());
|
Chris@0
|
720 }
|
Chris@0
|
721 }
|
Chris@0
|
722 }
|
Chris@0
|
723
|
Chris@0
|
724 void FeatureExtractionManager::finish()
|
Chris@0
|
725 {
|
Chris@0
|
726 for (PluginMap::iterator pi = m_plugins.begin();
|
Chris@0
|
727 pi != m_plugins.end(); ++pi) {
|
Chris@0
|
728
|
Chris@0
|
729 for (TransformWriterMap::iterator ti = pi->second.begin();
|
Chris@0
|
730 ti != pi->second.end(); ++ti) {
|
Chris@0
|
731
|
Chris@0
|
732 vector<FeatureWriter *> &writers = ti->second;
|
Chris@0
|
733
|
Chris@0
|
734 for (int i = 0; i < (int)writers.size(); ++i) {
|
Chris@0
|
735 writers[i]->flush();
|
Chris@0
|
736 writers[i]->finish();
|
Chris@0
|
737 }
|
Chris@0
|
738 }
|
Chris@0
|
739 }
|
Chris@0
|
740 }
|
Chris@0
|
741
|
Chris@0
|
742 void FeatureExtractionManager::print(Transform transform) const
|
Chris@0
|
743 {
|
Chris@0
|
744 QString qs;
|
Chris@0
|
745 QTextStream qts(&qs);
|
Chris@0
|
746 transform.toXml(qts);
|
Chris@0
|
747 cerr << qs.toStdString() << endl;
|
Chris@0
|
748 }
|