Chris@49
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
Chris@0
|
2
|
Chris@0
|
3 /*
|
Chris@52
|
4 Sonic Visualiser
|
Chris@52
|
5 An audio file viewer and annotation editor.
|
Chris@52
|
6 Centre for Digital Music, Queen Mary, University of London.
|
Chris@52
|
7 This file copyright 2006 Chris Cannam.
|
Chris@0
|
8
|
Chris@52
|
9 This program is free software; you can redistribute it and/or
|
Chris@52
|
10 modify it under the terms of the GNU General Public License as
|
Chris@52
|
11 published by the Free Software Foundation; either version 2 of the
|
Chris@52
|
12 License, or (at your option) any later version. See the file
|
Chris@52
|
13 COPYING included with this distribution for more information.
|
Chris@0
|
14 */
|
Chris@0
|
15
|
Chris@0
|
16 #include "FeatureExtractionPluginTransform.h"
|
Chris@0
|
17
|
Chris@0
|
18 #include "plugin/FeatureExtractionPluginFactory.h"
|
Chris@66
|
19 #include "plugin/PluginXml.h"
|
Chris@66
|
20 #include "vamp-sdk/Plugin.h"
|
Chris@0
|
21
|
Chris@0
|
22 #include "base/Model.h"
|
Chris@67
|
23 #include "base/Window.h"
|
Chris@0
|
24 #include "model/SparseOneDimensionalModel.h"
|
Chris@0
|
25 #include "model/SparseTimeValueModel.h"
|
Chris@0
|
26 #include "model/DenseThreeDimensionalModel.h"
|
Chris@0
|
27 #include "model/DenseTimeValueModel.h"
|
Chris@115
|
28 #include "model/NoteModel.h"
|
Chris@0
|
29
|
Chris@128
|
30 #include "fileio/FFTDataServer.h"
|
Chris@128
|
31
|
Chris@67
|
32 #include <fftw3.h>
|
Chris@67
|
33
|
Chris@0
|
34 #include <iostream>
|
Chris@0
|
35
|
Chris@0
|
36 FeatureExtractionPluginTransform::FeatureExtractionPluginTransform(Model *inputModel,
|
Chris@0
|
37 QString pluginId,
|
Chris@64
|
38 int channel,
|
Chris@56
|
39 QString configurationXml,
|
Chris@0
|
40 QString outputName) :
|
Chris@0
|
41 Transform(inputModel),
|
Chris@0
|
42 m_plugin(0),
|
Chris@64
|
43 m_channel(channel),
|
Chris@68
|
44 m_stepSize(0),
|
Chris@68
|
45 m_blockSize(0),
|
Chris@0
|
46 m_descriptor(0),
|
Chris@0
|
47 m_outputFeatureNo(0)
|
Chris@0
|
48 {
|
Chris@117
|
49 // std::cerr << "FeatureExtractionPluginTransform::FeatureExtractionPluginTransform: plugin " << pluginId.toStdString() << ", outputName " << outputName.toStdString() << std::endl;
|
Chris@0
|
50
|
Chris@0
|
51 FeatureExtractionPluginFactory *factory =
|
Chris@0
|
52 FeatureExtractionPluginFactory::instanceFor(pluginId);
|
Chris@0
|
53
|
Chris@0
|
54 if (!factory) {
|
Chris@0
|
55 std::cerr << "FeatureExtractionPluginTransform: No factory available for plugin id \""
|
Chris@0
|
56 << pluginId.toStdString() << "\"" << std::endl;
|
Chris@0
|
57 return;
|
Chris@0
|
58 }
|
Chris@0
|
59
|
Chris@0
|
60 m_plugin = factory->instantiatePlugin(pluginId, m_input->getSampleRate());
|
Chris@0
|
61
|
Chris@0
|
62 if (!m_plugin) {
|
Chris@0
|
63 std::cerr << "FeatureExtractionPluginTransform: Failed to instantiate plugin \""
|
Chris@0
|
64 << pluginId.toStdString() << "\"" << std::endl;
|
Chris@0
|
65 return;
|
Chris@0
|
66 }
|
Chris@0
|
67
|
Chris@56
|
68 if (configurationXml != "") {
|
Chris@66
|
69 PluginXml(m_plugin).setParametersFromXml(configurationXml);
|
Chris@56
|
70 }
|
Chris@56
|
71
|
Chris@68
|
72 m_blockSize = m_plugin->getPreferredBlockSize();
|
Chris@68
|
73 m_stepSize = m_plugin->getPreferredStepSize();
|
Chris@68
|
74
|
Chris@68
|
75 if (m_blockSize == 0) m_blockSize = 1024; //!!! todo: ask user
|
Chris@68
|
76 if (m_stepSize == 0) m_stepSize = m_blockSize; //!!! likewise
|
Chris@68
|
77
|
Chris@114
|
78 DenseTimeValueModel *input = getInput();
|
Chris@114
|
79 if (!input) return;
|
Chris@114
|
80
|
Chris@114
|
81 size_t channelCount = input->getChannelCount();
|
Chris@114
|
82 if (m_plugin->getMaxChannelCount() < channelCount) {
|
Chris@114
|
83 channelCount = 1;
|
Chris@114
|
84 }
|
Chris@114
|
85 if (m_plugin->getMinChannelCount() > channelCount) {
|
Chris@114
|
86 std::cerr << "FeatureExtractionPluginTransform:: "
|
Chris@114
|
87 << "Can't provide enough channels to plugin (plugin min "
|
Chris@114
|
88 << m_plugin->getMinChannelCount() << ", max "
|
Chris@114
|
89 << m_plugin->getMaxChannelCount() << ", input model has "
|
Chris@114
|
90 << input->getChannelCount() << ")" << std::endl;
|
Chris@114
|
91 return;
|
Chris@114
|
92 }
|
Chris@114
|
93
|
Chris@114
|
94 if (!m_plugin->initialise(channelCount, m_stepSize, m_blockSize)) {
|
Chris@114
|
95 std::cerr << "FeatureExtractionPluginTransform: Plugin "
|
Chris@114
|
96 << m_plugin->getName() << " failed to initialise!" << std::endl;
|
Chris@114
|
97 return;
|
Chris@114
|
98 }
|
Chris@114
|
99
|
Chris@74
|
100 Vamp::Plugin::OutputList outputs = m_plugin->getOutputDescriptors();
|
Chris@0
|
101
|
Chris@0
|
102 if (outputs.empty()) {
|
Chris@0
|
103 std::cerr << "FeatureExtractionPluginTransform: Plugin \""
|
Chris@0
|
104 << pluginId.toStdString() << "\" has no outputs" << std::endl;
|
Chris@0
|
105 return;
|
Chris@0
|
106 }
|
Chris@0
|
107
|
Chris@0
|
108 for (size_t i = 0; i < outputs.size(); ++i) {
|
Chris@0
|
109 if (outputName == "" || outputs[i].name == outputName.toStdString()) {
|
Chris@0
|
110 m_outputFeatureNo = i;
|
Chris@66
|
111 m_descriptor = new Vamp::Plugin::OutputDescriptor
|
Chris@0
|
112 (outputs[i]);
|
Chris@0
|
113 break;
|
Chris@0
|
114 }
|
Chris@0
|
115 }
|
Chris@0
|
116
|
Chris@0
|
117 if (!m_descriptor) {
|
Chris@0
|
118 std::cerr << "FeatureExtractionPluginTransform: Plugin \""
|
Chris@0
|
119 << pluginId.toStdString() << "\" has no output named \""
|
Chris@0
|
120 << outputName.toStdString() << "\"" << std::endl;
|
Chris@0
|
121 return;
|
Chris@0
|
122 }
|
Chris@0
|
123
|
Chris@117
|
124 // std::cerr << "FeatureExtractionPluginTransform: output sample type "
|
Chris@117
|
125 // << m_descriptor->sampleType << std::endl;
|
Chris@0
|
126
|
Chris@70
|
127 int binCount = 1;
|
Chris@0
|
128 float minValue = 0.0, maxValue = 0.0;
|
Chris@0
|
129
|
Chris@70
|
130 if (m_descriptor->hasFixedBinCount) {
|
Chris@70
|
131 binCount = m_descriptor->binCount;
|
Chris@0
|
132 }
|
Chris@0
|
133
|
Chris@117
|
134 // std::cerr << "FeatureExtractionPluginTransform: output bin count "
|
Chris@117
|
135 // << binCount << std::endl;
|
Chris@114
|
136
|
Chris@70
|
137 if (binCount > 0 && m_descriptor->hasKnownExtents) {
|
Chris@0
|
138 minValue = m_descriptor->minValue;
|
Chris@0
|
139 maxValue = m_descriptor->maxValue;
|
Chris@0
|
140 }
|
Chris@0
|
141
|
Chris@0
|
142 size_t modelRate = m_input->getSampleRate();
|
Chris@0
|
143 size_t modelResolution = 1;
|
Chris@0
|
144
|
Chris@0
|
145 switch (m_descriptor->sampleType) {
|
Chris@0
|
146
|
Chris@66
|
147 case Vamp::Plugin::OutputDescriptor::VariableSampleRate:
|
Chris@0
|
148 if (m_descriptor->sampleRate != 0.0) {
|
Chris@0
|
149 modelResolution = size_t(modelRate / m_descriptor->sampleRate + 0.001);
|
Chris@0
|
150 }
|
Chris@0
|
151 break;
|
Chris@0
|
152
|
Chris@66
|
153 case Vamp::Plugin::OutputDescriptor::OneSamplePerStep:
|
Chris@72
|
154 modelResolution = m_stepSize;
|
Chris@0
|
155 break;
|
Chris@0
|
156
|
Chris@66
|
157 case Vamp::Plugin::OutputDescriptor::FixedSampleRate:
|
Chris@73
|
158 modelRate = size_t(m_descriptor->sampleRate + 0.001);
|
Chris@0
|
159 break;
|
Chris@0
|
160 }
|
Chris@0
|
161
|
Chris@70
|
162 if (binCount == 0) {
|
Chris@0
|
163
|
Chris@20
|
164 m_output = new SparseOneDimensionalModel(modelRate, modelResolution,
|
Chris@20
|
165 false);
|
Chris@0
|
166
|
Chris@115
|
167 } else if (binCount == 1) {
|
Chris@0
|
168
|
Chris@115
|
169 SparseTimeValueModel *model = new SparseTimeValueModel
|
Chris@115
|
170 (modelRate, modelResolution, minValue, maxValue, false);
|
Chris@115
|
171 model->setScaleUnits(outputs[m_outputFeatureNo].unit.c_str());
|
Chris@115
|
172
|
Chris@115
|
173 m_output = model;
|
Chris@115
|
174
|
Chris@115
|
175 } else if (m_descriptor->sampleType ==
|
Chris@66
|
176 Vamp::Plugin::OutputDescriptor::VariableSampleRate) {
|
Chris@115
|
177
|
Chris@115
|
178 // We don't have a sparse 3D model, so interpret this as a
|
Chris@115
|
179 // note model. There's nothing to define which values to use
|
Chris@115
|
180 // as which parameters of the note -- for the moment let's
|
Chris@115
|
181 // treat the first as pitch, second as duration in frames,
|
Chris@115
|
182 // third (if present) as velocity. (Our note model doesn't
|
Chris@115
|
183 // yet store velocity.)
|
Chris@115
|
184 //!!! todo: ask the user!
|
Chris@0
|
185
|
Chris@115
|
186 NoteModel *model = new NoteModel
|
Chris@63
|
187 (modelRate, modelResolution, minValue, maxValue, false);
|
Chris@63
|
188 model->setScaleUnits(outputs[m_outputFeatureNo].unit.c_str());
|
Chris@63
|
189
|
Chris@63
|
190 m_output = model;
|
Chris@0
|
191
|
Chris@0
|
192 } else {
|
Chris@0
|
193
|
Chris@0
|
194 m_output = new DenseThreeDimensionalModel(modelRate, modelResolution,
|
Chris@70
|
195 binCount, false);
|
Chris@20
|
196
|
Chris@70
|
197 if (!m_descriptor->binNames.empty()) {
|
Chris@20
|
198 std::vector<QString> names;
|
Chris@70
|
199 for (size_t i = 0; i < m_descriptor->binNames.size(); ++i) {
|
Chris@70
|
200 names.push_back(m_descriptor->binNames[i].c_str());
|
Chris@20
|
201 }
|
Chris@20
|
202 (dynamic_cast<DenseThreeDimensionalModel *>(m_output))
|
Chris@20
|
203 ->setBinNames(names);
|
Chris@20
|
204 }
|
Chris@0
|
205 }
|
Chris@0
|
206 }
|
Chris@0
|
207
|
Chris@0
|
208 FeatureExtractionPluginTransform::~FeatureExtractionPluginTransform()
|
Chris@0
|
209 {
|
Chris@0
|
210 delete m_plugin;
|
Chris@0
|
211 delete m_descriptor;
|
Chris@0
|
212 }
|
Chris@0
|
213
|
Chris@0
|
214 DenseTimeValueModel *
|
Chris@0
|
215 FeatureExtractionPluginTransform::getInput()
|
Chris@0
|
216 {
|
Chris@0
|
217 DenseTimeValueModel *dtvm =
|
Chris@0
|
218 dynamic_cast<DenseTimeValueModel *>(getInputModel());
|
Chris@0
|
219 if (!dtvm) {
|
Chris@0
|
220 std::cerr << "FeatureExtractionPluginTransform::getInput: WARNING: Input model is not conformable to DenseTimeValueModel" << std::endl;
|
Chris@0
|
221 }
|
Chris@0
|
222 return dtvm;
|
Chris@0
|
223 }
|
Chris@0
|
224
|
Chris@0
|
225 void
|
Chris@0
|
226 FeatureExtractionPluginTransform::run()
|
Chris@0
|
227 {
|
Chris@0
|
228 DenseTimeValueModel *input = getInput();
|
Chris@0
|
229 if (!input) return;
|
Chris@0
|
230
|
Chris@0
|
231 if (!m_output) return;
|
Chris@0
|
232
|
Chris@114
|
233 size_t sampleRate = m_input->getSampleRate();
|
Chris@114
|
234
|
Chris@0
|
235 size_t channelCount = input->getChannelCount();
|
Chris@0
|
236 if (m_plugin->getMaxChannelCount() < channelCount) {
|
Chris@0
|
237 channelCount = 1;
|
Chris@0
|
238 }
|
Chris@0
|
239
|
Chris@0
|
240 float **buffers = new float*[channelCount];
|
Chris@0
|
241 for (size_t ch = 0; ch < channelCount; ++ch) {
|
Chris@68
|
242 buffers[ch] = new float[m_blockSize];
|
Chris@0
|
243 }
|
Chris@0
|
244
|
Chris@128
|
245 /*!!!
|
Chris@122
|
246 float *fftInput = 0;
|
Chris@122
|
247 fftwf_complex *fftOutput = 0;
|
Chris@122
|
248 fftwf_plan fftPlan = 0;
|
Chris@122
|
249 Window<float> windower(HanningWindow, m_blockSize);
|
Chris@67
|
250
|
Chris@67
|
251 if (m_plugin->getInputDomain() == Vamp::Plugin::FrequencyDomain) {
|
Chris@67
|
252
|
Chris@122
|
253 fftInput = (float *)fftwf_malloc(m_blockSize * sizeof(double));
|
Chris@122
|
254 fftOutput = (fftwf_complex *)fftwf_malloc(m_blockSize * sizeof(fftwf_complex));
|
Chris@122
|
255 fftPlan = fftwf_plan_dft_r2c_1d(m_blockSize, fftInput, fftOutput,
|
Chris@122
|
256 FFTW_ESTIMATE);
|
Chris@67
|
257 if (!fftPlan) {
|
Chris@67
|
258 std::cerr << "ERROR: FeatureExtractionPluginTransform::run(): fftw_plan failed! Results will be garbage" << std::endl;
|
Chris@67
|
259 }
|
Chris@67
|
260 }
|
Chris@128
|
261 */
|
Chris@128
|
262
|
Chris@128
|
263 bool frequencyDomain = (m_plugin->getInputDomain() ==
|
Chris@128
|
264 Vamp::Plugin::FrequencyDomain);
|
Chris@128
|
265 std::vector<FFTDataServer *> fftServers;
|
Chris@128
|
266
|
Chris@131
|
267 //!!! use fuzzy fft server matching!
|
Chris@131
|
268
|
Chris@128
|
269 if (frequencyDomain) {
|
Chris@128
|
270 for (size_t ch = 0; ch < channelCount; ++ch) {
|
Chris@128
|
271 fftServers.push_back(FFTDataServer::getInstance
|
Chris@128
|
272 (getInput(),
|
Chris@128
|
273 channelCount == 1 ? m_channel : ch,
|
Chris@128
|
274 HanningWindow,
|
Chris@128
|
275 m_blockSize,
|
Chris@128
|
276 m_stepSize,
|
Chris@128
|
277 m_blockSize,
|
Chris@128
|
278 false));
|
Chris@128
|
279 }
|
Chris@128
|
280 }
|
Chris@67
|
281
|
Chris@70
|
282 long startFrame = m_input->getStartFrame();
|
Chris@70
|
283 long endFrame = m_input->getEndFrame();
|
Chris@70
|
284 long blockFrame = startFrame;
|
Chris@0
|
285
|
Chris@70
|
286 long prevCompletion = 0;
|
Chris@0
|
287
|
Chris@70
|
288 while (1) {
|
Chris@70
|
289
|
Chris@128
|
290 if (frequencyDomain) {
|
Chris@73
|
291 if (blockFrame - int(m_blockSize)/2 > endFrame) break;
|
Chris@70
|
292 } else {
|
Chris@70
|
293 if (blockFrame >= endFrame) break;
|
Chris@70
|
294 }
|
Chris@0
|
295
|
Chris@0
|
296 // std::cerr << "FeatureExtractionPluginTransform::run: blockFrame "
|
Chris@0
|
297 // << blockFrame << std::endl;
|
Chris@0
|
298
|
Chris@70
|
299 long completion =
|
Chris@68
|
300 (((blockFrame - startFrame) / m_stepSize) * 99) /
|
Chris@68
|
301 ( (endFrame - startFrame) / m_stepSize);
|
Chris@0
|
302
|
Chris@0
|
303 // channelCount is either m_input->channelCount or 1
|
Chris@0
|
304
|
Chris@73
|
305 for (size_t ch = 0; ch < channelCount; ++ch) {
|
Chris@128
|
306 //!!! if (fftPlan) {
|
Chris@128
|
307 // getFrames(ch, channelCount,
|
Chris@128
|
308 // blockFrame - m_blockSize/2, m_blockSize, buffers[ch]);
|
Chris@128
|
309
|
Chris@128
|
310 if (frequencyDomain) {
|
Chris@128
|
311 int column = (blockFrame - startFrame) / m_stepSize;
|
Chris@128
|
312 for (size_t i = 0; i < m_blockSize/2; ++i) {
|
Chris@128
|
313 fftServers[ch]->getValuesAt
|
Chris@128
|
314 (column, i, buffers[ch][i*2], buffers[ch][i*2+1]);
|
Chris@128
|
315 }
|
Chris@70
|
316 } else {
|
Chris@70
|
317 getFrames(ch, channelCount,
|
Chris@128
|
318 blockFrame, m_blockSize, buffers[ch]);
|
Chris@70
|
319 }
|
Chris@70
|
320 }
|
Chris@128
|
321 /*!!!
|
Chris@67
|
322 if (fftPlan) {
|
Chris@73
|
323 for (size_t ch = 0; ch < channelCount; ++ch) {
|
Chris@73
|
324 for (size_t i = 0; i < m_blockSize; ++i) {
|
Chris@67
|
325 fftInput[i] = buffers[ch][i];
|
Chris@67
|
326 }
|
Chris@67
|
327 windower.cut(fftInput);
|
Chris@73
|
328 for (size_t i = 0; i < m_blockSize/2; ++i) {
|
Chris@122
|
329 float temp = fftInput[i];
|
Chris@68
|
330 fftInput[i] = fftInput[i + m_blockSize/2];
|
Chris@68
|
331 fftInput[i + m_blockSize/2] = temp;
|
Chris@67
|
332 }
|
Chris@122
|
333 fftwf_execute(fftPlan);
|
Chris@73
|
334 for (size_t i = 0; i < m_blockSize/2; ++i) {
|
Chris@67
|
335 buffers[ch][i*2] = fftOutput[i][0];
|
Chris@67
|
336 buffers[ch][i*2 + 1] = fftOutput[i][1];
|
Chris@67
|
337 }
|
Chris@67
|
338 }
|
Chris@67
|
339 }
|
Chris@128
|
340 */
|
Chris@66
|
341 Vamp::Plugin::FeatureSet features = m_plugin->process
|
Chris@66
|
342 (buffers, Vamp::RealTime::frame2RealTime(blockFrame, sampleRate));
|
Chris@0
|
343
|
Chris@0
|
344 for (size_t fi = 0; fi < features[m_outputFeatureNo].size(); ++fi) {
|
Chris@66
|
345 Vamp::Plugin::Feature feature =
|
Chris@0
|
346 features[m_outputFeatureNo][fi];
|
Chris@0
|
347 addFeature(blockFrame, feature);
|
Chris@0
|
348 }
|
Chris@0
|
349
|
Chris@0
|
350 if (blockFrame == startFrame || completion > prevCompletion) {
|
Chris@0
|
351 setCompletion(completion);
|
Chris@0
|
352 prevCompletion = completion;
|
Chris@0
|
353 }
|
Chris@0
|
354
|
Chris@68
|
355 blockFrame += m_stepSize;
|
Chris@0
|
356 }
|
Chris@128
|
357 /*!!!
|
Chris@67
|
358 if (fftPlan) {
|
Chris@122
|
359 fftwf_destroy_plan(fftPlan);
|
Chris@122
|
360 fftwf_free(fftInput);
|
Chris@122
|
361 fftwf_free(fftOutput);
|
Chris@67
|
362 }
|
Chris@128
|
363 */
|
Chris@66
|
364 Vamp::Plugin::FeatureSet features = m_plugin->getRemainingFeatures();
|
Chris@0
|
365
|
Chris@0
|
366 for (size_t fi = 0; fi < features[m_outputFeatureNo].size(); ++fi) {
|
Chris@66
|
367 Vamp::Plugin::Feature feature =
|
Chris@0
|
368 features[m_outputFeatureNo][fi];
|
Chris@0
|
369 addFeature(blockFrame, feature);
|
Chris@0
|
370 }
|
Chris@0
|
371
|
Chris@0
|
372 setCompletion(100);
|
Chris@0
|
373 }
|
Chris@0
|
374
|
Chris@70
|
375 void
|
Chris@70
|
376 FeatureExtractionPluginTransform::getFrames(int channel, int channelCount,
|
Chris@70
|
377 long startFrame, long size,
|
Chris@70
|
378 float *buffer)
|
Chris@70
|
379 {
|
Chris@70
|
380 long offset = 0;
|
Chris@70
|
381
|
Chris@70
|
382 if (startFrame < 0) {
|
Chris@70
|
383 for (int i = 0; i < size && startFrame + i < 0; ++i) {
|
Chris@70
|
384 buffer[i] = 0.0f;
|
Chris@70
|
385 }
|
Chris@70
|
386 offset = -startFrame;
|
Chris@70
|
387 size -= offset;
|
Chris@70
|
388 if (size <= 0) return;
|
Chris@70
|
389 startFrame = 0;
|
Chris@70
|
390 }
|
Chris@70
|
391
|
Chris@73
|
392 long got = getInput()->getValues
|
Chris@70
|
393 ((channelCount == 1 ? m_channel : channel),
|
Chris@70
|
394 startFrame, startFrame + size, buffer + offset);
|
Chris@70
|
395
|
Chris@70
|
396 while (got < size) {
|
Chris@70
|
397 buffer[offset + got] = 0.0;
|
Chris@70
|
398 ++got;
|
Chris@70
|
399 }
|
Chris@74
|
400
|
Chris@74
|
401 if (m_channel == -1 && channelCount == 1 &&
|
Chris@74
|
402 getInput()->getChannelCount() > 1) {
|
Chris@74
|
403 // use mean instead of sum, as plugin input
|
Chris@74
|
404 int cc = getInput()->getChannelCount();
|
Chris@74
|
405 for (long i = 0; i < size; ++i) {
|
Chris@74
|
406 buffer[i] /= cc;
|
Chris@74
|
407 }
|
Chris@74
|
408 }
|
Chris@70
|
409 }
|
Chris@0
|
410
|
Chris@0
|
411 void
|
Chris@0
|
412 FeatureExtractionPluginTransform::addFeature(size_t blockFrame,
|
Chris@66
|
413 const Vamp::Plugin::Feature &feature)
|
Chris@0
|
414 {
|
Chris@0
|
415 size_t inputRate = m_input->getSampleRate();
|
Chris@0
|
416
|
Chris@0
|
417 // std::cerr << "FeatureExtractionPluginTransform::addFeature("
|
Chris@0
|
418 // << blockFrame << ")" << std::endl;
|
Chris@0
|
419
|
Chris@70
|
420 int binCount = 1;
|
Chris@70
|
421 if (m_descriptor->hasFixedBinCount) {
|
Chris@70
|
422 binCount = m_descriptor->binCount;
|
Chris@0
|
423 }
|
Chris@0
|
424
|
Chris@0
|
425 size_t frame = blockFrame;
|
Chris@0
|
426
|
Chris@0
|
427 if (m_descriptor->sampleType ==
|
Chris@66
|
428 Vamp::Plugin::OutputDescriptor::VariableSampleRate) {
|
Chris@0
|
429
|
Chris@0
|
430 if (!feature.hasTimestamp) {
|
Chris@0
|
431 std::cerr
|
Chris@0
|
432 << "WARNING: FeatureExtractionPluginTransform::addFeature: "
|
Chris@0
|
433 << "Feature has variable sample rate but no timestamp!"
|
Chris@0
|
434 << std::endl;
|
Chris@0
|
435 return;
|
Chris@0
|
436 } else {
|
Chris@66
|
437 frame = Vamp::RealTime::realTime2Frame(feature.timestamp, inputRate);
|
Chris@0
|
438 }
|
Chris@0
|
439
|
Chris@0
|
440 } else if (m_descriptor->sampleType ==
|
Chris@66
|
441 Vamp::Plugin::OutputDescriptor::FixedSampleRate) {
|
Chris@0
|
442
|
Chris@0
|
443 if (feature.hasTimestamp) {
|
Chris@0
|
444 //!!! warning: sampleRate may be non-integral
|
Chris@66
|
445 frame = Vamp::RealTime::realTime2Frame(feature.timestamp,
|
Chris@66
|
446 m_descriptor->sampleRate);
|
Chris@0
|
447 } else {
|
Chris@0
|
448 frame = m_output->getEndFrame() + 1;
|
Chris@0
|
449 }
|
Chris@0
|
450 }
|
Chris@0
|
451
|
Chris@70
|
452 if (binCount == 0) {
|
Chris@0
|
453
|
Chris@0
|
454 SparseOneDimensionalModel *model = getOutput<SparseOneDimensionalModel>();
|
Chris@0
|
455 if (!model) return;
|
Chris@0
|
456 model->addPoint(SparseOneDimensionalModel::Point(frame, feature.label.c_str()));
|
Chris@0
|
457
|
Chris@115
|
458 } else if (binCount == 1) {
|
Chris@0
|
459
|
Chris@0
|
460 float value = 0.0;
|
Chris@0
|
461 if (feature.values.size() > 0) value = feature.values[0];
|
Chris@0
|
462
|
Chris@0
|
463 SparseTimeValueModel *model = getOutput<SparseTimeValueModel>();
|
Chris@0
|
464 if (!model) return;
|
Chris@0
|
465 model->addPoint(SparseTimeValueModel::Point(frame, value, feature.label.c_str()));
|
Chris@115
|
466
|
Chris@115
|
467 } else if (m_descriptor->sampleType ==
|
Chris@115
|
468 Vamp::Plugin::OutputDescriptor::VariableSampleRate) {
|
Chris@115
|
469
|
Chris@115
|
470 float pitch = 0.0;
|
Chris@115
|
471 if (feature.values.size() > 0) pitch = feature.values[0];
|
Chris@115
|
472
|
Chris@115
|
473 float duration = 1;
|
Chris@115
|
474 if (feature.values.size() > 1) duration = feature.values[1];
|
Chris@115
|
475
|
Chris@115
|
476 float velocity = 100;
|
Chris@115
|
477 if (feature.values.size() > 2) velocity = feature.values[2];
|
Chris@115
|
478
|
Chris@115
|
479 NoteModel *model = getOutput<NoteModel>();
|
Chris@115
|
480 if (!model) return;
|
Chris@115
|
481
|
Chris@115
|
482 model->addPoint(NoteModel::Point(frame, pitch, duration, feature.label.c_str()));
|
Chris@0
|
483
|
Chris@0
|
484 } else {
|
Chris@0
|
485
|
Chris@0
|
486 DenseThreeDimensionalModel::BinValueSet values = feature.values;
|
Chris@0
|
487
|
Chris@0
|
488 DenseThreeDimensionalModel *model = getOutput<DenseThreeDimensionalModel>();
|
Chris@0
|
489 if (!model) return;
|
Chris@0
|
490
|
Chris@0
|
491 model->setBinValues(frame, values);
|
Chris@0
|
492 }
|
Chris@0
|
493 }
|
Chris@0
|
494
|
Chris@0
|
495 void
|
Chris@0
|
496 FeatureExtractionPluginTransform::setCompletion(int completion)
|
Chris@0
|
497 {
|
Chris@70
|
498 int binCount = 1;
|
Chris@70
|
499 if (m_descriptor->hasFixedBinCount) {
|
Chris@70
|
500 binCount = m_descriptor->binCount;
|
Chris@0
|
501 }
|
Chris@0
|
502
|
Chris@70
|
503 if (binCount == 0) {
|
Chris@0
|
504
|
Chris@0
|
505 SparseOneDimensionalModel *model = getOutput<SparseOneDimensionalModel>();
|
Chris@0
|
506 if (!model) return;
|
Chris@0
|
507 model->setCompletion(completion);
|
Chris@0
|
508
|
Chris@115
|
509 } else if (binCount == 1) {
|
Chris@115
|
510
|
Chris@115
|
511 SparseTimeValueModel *model = getOutput<SparseTimeValueModel>();
|
Chris@115
|
512 if (!model) return;
|
Chris@115
|
513 model->setCompletion(completion);
|
Chris@115
|
514
|
Chris@115
|
515 } else if (m_descriptor->sampleType ==
|
Chris@66
|
516 Vamp::Plugin::OutputDescriptor::VariableSampleRate) {
|
Chris@0
|
517
|
Chris@115
|
518 NoteModel *model = getOutput<NoteModel>();
|
Chris@0
|
519 if (!model) return;
|
Chris@0
|
520 model->setCompletion(completion);
|
Chris@0
|
521
|
Chris@0
|
522 } else {
|
Chris@0
|
523
|
Chris@19
|
524 DenseThreeDimensionalModel *model = getOutput<DenseThreeDimensionalModel>();
|
Chris@19
|
525 if (!model) return;
|
Chris@19
|
526 model->setCompletion(completion);
|
Chris@0
|
527 }
|
Chris@0
|
528 }
|
Chris@0
|
529
|