Chris@145
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
Chris@145
|
2
|
Chris@145
|
3 /*
|
Chris@145
|
4 Sonic Annotator
|
Chris@145
|
5 A utility for batch feature extraction from audio files.
|
Chris@145
|
6 Mark Levy, Chris Sutton and Chris Cannam, Queen Mary, University of London.
|
Chris@145
|
7 Copyright 2007-2014 QMUL.
|
Chris@145
|
8
|
Chris@145
|
9 This program is free software; you can redistribute it and/or
|
Chris@145
|
10 modify it under the terms of the GNU General Public License as
|
Chris@145
|
11 published by the Free Software Foundation; either version 2 of the
|
Chris@145
|
12 License, or (at your option) any later version. See the file
|
Chris@145
|
13 COPYING included with this distribution for more information.
|
Chris@145
|
14 */
|
Chris@145
|
15
|
Chris@145
|
16 #include "JAMSFeatureWriter.h"
|
Chris@145
|
17
|
Chris@145
|
18 using namespace std;
|
Chris@145
|
19 using Vamp::Plugin;
|
Chris@145
|
20 using Vamp::PluginBase;
|
Chris@145
|
21
|
Chris@145
|
22 #include "base/Exceptions.h"
|
Chris@145
|
23 #include "rdf/PluginRDFIndexer.h"
|
Chris@145
|
24
|
Chris@162
|
25 #include "version.h"
|
Chris@162
|
26
|
Chris@145
|
27 JAMSFeatureWriter::JAMSFeatureWriter() :
|
Chris@145
|
28 FileFeatureWriter(SupportOneFilePerTrackTransform |
|
Chris@145
|
29 SupportOneFilePerTrack |
|
Chris@152
|
30 SupportOneFileTotal |
|
Chris@145
|
31 SupportStdOut,
|
Chris@145
|
32 "json"),
|
Chris@145
|
33 m_network(false),
|
Chris@145
|
34 m_networkRetrieved(false)
|
Chris@145
|
35 {
|
Chris@145
|
36 }
|
Chris@145
|
37
|
Chris@145
|
38 JAMSFeatureWriter::~JAMSFeatureWriter()
|
Chris@145
|
39 {
|
Chris@145
|
40 }
|
Chris@145
|
41
|
Chris@145
|
42 string
|
Chris@145
|
43 JAMSFeatureWriter::getDescription() const
|
Chris@145
|
44 {
|
Chris@145
|
45 return "Write features to JSON files in JAMS (JSON Annotated Music Specification) format.";
|
Chris@145
|
46 }
|
Chris@145
|
47
|
Chris@145
|
48 JAMSFeatureWriter::ParameterList
|
Chris@145
|
49 JAMSFeatureWriter::getSupportedParameters() const
|
Chris@145
|
50 {
|
Chris@145
|
51 ParameterList pl = FileFeatureWriter::getSupportedParameters();
|
Chris@145
|
52 Parameter p;
|
Chris@145
|
53
|
Chris@145
|
54 p.name = "network";
|
Chris@145
|
55 p.description = "Attempt to retrieve RDF descriptions of plugins from network, if not available locally";
|
Chris@145
|
56 p.hasArg = false;
|
Chris@145
|
57 pl.push_back(p);
|
Chris@145
|
58
|
Chris@145
|
59 return pl;
|
Chris@145
|
60 }
|
Chris@145
|
61
|
Chris@145
|
62 void
|
Chris@145
|
63 JAMSFeatureWriter::setParameters(map<string, string> ¶ms)
|
Chris@145
|
64 {
|
Chris@145
|
65 FileFeatureWriter::setParameters(params);
|
Chris@145
|
66
|
Chris@145
|
67 for (map<string, string>::iterator i = params.begin();
|
Chris@145
|
68 i != params.end(); ++i) {
|
Chris@145
|
69 if (i->first == "network") {
|
Chris@145
|
70 m_network = true;
|
Chris@145
|
71 }
|
Chris@145
|
72 }
|
Chris@145
|
73 }
|
Chris@145
|
74
|
Chris@145
|
75 void
|
Chris@145
|
76 JAMSFeatureWriter::setTrackMetadata(QString trackId, TrackMetadata metadata)
|
Chris@145
|
77 {
|
Chris@145
|
78 QString json
|
Chris@162
|
79 ("\n\"file_metadata\": {\n"
|
Chris@162
|
80 " \"artist\": \"%1\",\n"
|
Chris@162
|
81 " \"title\": \"%2\"\n},\n");
|
Chris@145
|
82 m_metadata[trackId] = json.arg(metadata.maker).arg(metadata.title);
|
Chris@152
|
83 cerr << "setTrackMetadata: metadata is: " << m_metadata[trackId] << endl;
|
Chris@145
|
84 }
|
Chris@145
|
85
|
Chris@153
|
86 static double
|
Chris@153
|
87 realTime2Sec(const Vamp::RealTime &r)
|
Chris@153
|
88 {
|
Chris@153
|
89 return r / Vamp::RealTime(1, 0);
|
Chris@153
|
90 }
|
Chris@153
|
91
|
Chris@145
|
92 void
|
Chris@145
|
93 JAMSFeatureWriter::write(QString trackId,
|
Chris@145
|
94 const Transform &transform,
|
Chris@145
|
95 const Plugin::OutputDescriptor& ,
|
Chris@145
|
96 const Plugin::FeatureList& features,
|
Chris@145
|
97 std::string /* summaryType */)
|
Chris@145
|
98 {
|
Chris@145
|
99 QString transformId = transform.getIdentifier();
|
Chris@145
|
100
|
Chris@145
|
101 QTextStream *sptr = getOutputStream(trackId, transformId);
|
Chris@145
|
102 if (!sptr) {
|
Chris@145
|
103 throw FailedToOpenOutputStream(trackId, transformId);
|
Chris@145
|
104 }
|
Chris@145
|
105
|
Chris@145
|
106 QTextStream &stream = *sptr;
|
Chris@145
|
107
|
Chris@152
|
108 TrackTransformPair tt(trackId, transformId);
|
Chris@152
|
109 TrackTransformPair targetKey = getFilenameKey(trackId, transformId);
|
Chris@152
|
110
|
Chris@152
|
111 if (m_startedTargets.find(targetKey) == m_startedTargets.end()) {
|
Chris@152
|
112 // Need to write track-level preamble
|
Chris@152
|
113 stream << "{" << m_metadata[trackId] << endl;
|
Chris@152
|
114 m_startedTargets.insert(targetKey);
|
Chris@152
|
115 }
|
Chris@152
|
116
|
Chris@153
|
117 bool justBegun = false;
|
Chris@153
|
118
|
Chris@152
|
119 if (m_data.find(tt) == m_data.end()) {
|
Chris@145
|
120
|
Chris@145
|
121 identifyTask(transform);
|
Chris@145
|
122
|
Chris@162
|
123 QString json
|
Chris@162
|
124 ("\"%1\": [ { \n"
|
Chris@162
|
125 " \"annotation_metadata\": {\n"
|
Chris@162
|
126 " \"annotation_tools\": \"Sonic Annotator v%2\",\n"
|
Chris@162
|
127 " \"data_source\": \"Automatic feature extraction\",\n"
|
Chris@165
|
128 " \"annotator\": {\n"
|
Chris@165
|
129 "%3"
|
Chris@165
|
130 " },\n"
|
Chris@162
|
131 " },\n"
|
Chris@162
|
132 " \"data\": [");
|
Chris@162
|
133 m_data[tt] = json
|
Chris@162
|
134 .arg(getTaskKey(m_tasks[transformId]))
|
Chris@162
|
135 .arg(RUNNER_VERSION)
|
Chris@165
|
136 .arg(writeTransformToObjectContents(transform));
|
Chris@153
|
137 justBegun = true;
|
Chris@145
|
138 }
|
Chris@145
|
139
|
Chris@153
|
140 QString d = m_data[tt];
|
Chris@153
|
141
|
Chris@145
|
142 for (int i = 0; i < int(features.size()); ++i) {
|
Chris@153
|
143
|
Chris@153
|
144 if (i > 0 || !justBegun) {
|
Chris@153
|
145 d += ",\n";
|
Chris@153
|
146 } else {
|
Chris@153
|
147 d += "\n";
|
Chris@153
|
148 }
|
Chris@153
|
149
|
Chris@153
|
150 d += " { ";
|
Chris@145
|
151
|
Chris@153
|
152 Plugin::Feature f(features[i]);
|
Chris@153
|
153
|
Chris@153
|
154 switch (m_tasks[transformId]) {
|
Chris@153
|
155
|
Chris@153
|
156 case ChordTask:
|
Chris@153
|
157 case SegmentTask:
|
Chris@153
|
158 case NoteTask:
|
Chris@153
|
159 case UnknownTask:
|
Chris@153
|
160 if (f.hasDuration) {
|
Chris@153
|
161 d += QString
|
Chris@153
|
162 ("\"start\": { \"value\": %1 }, "
|
Chris@153
|
163 "\"end\": { \"value\": %2 }")
|
Chris@153
|
164 .arg(realTime2Sec(f.timestamp))
|
Chris@153
|
165 .arg(realTime2Sec
|
Chris@153
|
166 (f.timestamp +
|
Chris@153
|
167 (f.hasDuration ? f.duration : Vamp::RealTime::zeroTime)));
|
Chris@153
|
168 break;
|
Chris@153
|
169 } else {
|
Chris@153
|
170 // don't break; fall through to simpler no-duration case
|
Chris@153
|
171 }
|
Chris@153
|
172
|
Chris@153
|
173 case BeatTask:
|
Chris@153
|
174 case KeyTask:
|
Chris@153
|
175 case OnsetTask:
|
Chris@153
|
176 d += QString("\"time\": { \"value\": %1 }")
|
Chris@153
|
177 .arg(realTime2Sec(f.timestamp));
|
Chris@153
|
178 break;
|
Chris@161
|
179
|
Chris@161
|
180 case MelodyTask:
|
Chris@161
|
181 case PitchTask:
|
Chris@161
|
182 //!!!
|
Chris@161
|
183 break;
|
Chris@153
|
184 }
|
Chris@153
|
185
|
Chris@153
|
186 if (f.label != "") {
|
Chris@153
|
187 d += QString(", \"label\": { \"value\": \"%2\" }")
|
Chris@153
|
188 .arg(f.label.c_str());
|
Chris@153
|
189 } else if (f.values.size() > 0) {
|
Chris@153
|
190 d += QString(", \"label\": { \"value\": \"%2\" }")
|
Chris@153
|
191 .arg(f.values[0]);
|
Chris@153
|
192 }
|
Chris@153
|
193
|
Chris@153
|
194 d += " }";
|
Chris@145
|
195 }
|
Chris@153
|
196
|
Chris@153
|
197 m_data[tt] = d;
|
Chris@145
|
198 }
|
Chris@145
|
199
|
Chris@145
|
200 void
|
Chris@152
|
201 JAMSFeatureWriter::finish()
|
Chris@152
|
202 {
|
Chris@152
|
203 cerr << "Finish called on " << this << endl;
|
Chris@152
|
204
|
Chris@152
|
205 set<QTextStream *> startedStreams;
|
Chris@152
|
206
|
Chris@152
|
207 for (DataMap::const_iterator i = m_data.begin();
|
Chris@152
|
208 i != m_data.end(); ++i) {
|
Chris@152
|
209
|
Chris@152
|
210 TrackTransformPair tt = i->first;
|
Chris@152
|
211 QString data = i->second;
|
Chris@152
|
212
|
Chris@152
|
213 QTextStream *sptr = getOutputStream(tt.first, tt.second);
|
Chris@152
|
214 if (!sptr) {
|
Chris@152
|
215 throw FailedToOpenOutputStream(tt.first, tt.second);
|
Chris@152
|
216 }
|
Chris@152
|
217
|
Chris@152
|
218 if (startedStreams.find(sptr) != startedStreams.end()) {
|
Chris@152
|
219 *sptr << "," << endl;
|
Chris@152
|
220 }
|
Chris@152
|
221 startedStreams.insert(sptr);
|
Chris@152
|
222
|
Chris@162
|
223 *sptr << data << "\n ]\n} ]";
|
Chris@152
|
224 }
|
Chris@152
|
225
|
Chris@152
|
226 for (FileStreamMap::const_iterator i = m_streams.begin();
|
Chris@152
|
227 i != m_streams.end(); ++i) {
|
Chris@152
|
228 *(i->second) << endl << "}" << endl;
|
Chris@152
|
229 }
|
Chris@152
|
230
|
Chris@152
|
231 m_data.clear();
|
Chris@152
|
232 m_startedTargets.clear();
|
Chris@152
|
233
|
Chris@152
|
234 FileFeatureWriter::finish();
|
Chris@152
|
235 }
|
Chris@152
|
236
|
Chris@152
|
237 void
|
Chris@145
|
238 JAMSFeatureWriter::loadRDFDescription(const Transform &transform)
|
Chris@145
|
239 {
|
Chris@145
|
240 QString pluginId = transform.getPluginIdentifier();
|
Chris@145
|
241 if (m_rdfDescriptions.find(pluginId) != m_rdfDescriptions.end()) return;
|
Chris@145
|
242
|
Chris@145
|
243 if (m_network && !m_networkRetrieved) {
|
Chris@145
|
244 PluginRDFIndexer::getInstance()->indexConfiguredURLs();
|
Chris@145
|
245 m_networkRetrieved = true;
|
Chris@145
|
246 }
|
Chris@145
|
247
|
Chris@145
|
248 m_rdfDescriptions[pluginId] = PluginRDFDescription(pluginId);
|
Chris@145
|
249
|
Chris@145
|
250 if (m_rdfDescriptions[pluginId].haveDescription()) {
|
Chris@145
|
251 cerr << "NOTE: Have RDF description for plugin ID \""
|
Chris@145
|
252 << pluginId << "\"" << endl;
|
Chris@145
|
253 } else {
|
Chris@145
|
254 cerr << "NOTE: No RDF description for plugin ID \""
|
Chris@145
|
255 << pluginId << "\"" << endl;
|
Chris@145
|
256 if (!m_network) {
|
Chris@145
|
257 cerr << " Consider using the --json-network option to retrieve plugin descriptions" << endl;
|
Chris@145
|
258 cerr << " from the network where possible." << endl;
|
Chris@145
|
259 }
|
Chris@145
|
260 }
|
Chris@145
|
261 }
|
Chris@145
|
262
|
Chris@145
|
263 void
|
Chris@145
|
264 JAMSFeatureWriter::identifyTask(const Transform &transform)
|
Chris@145
|
265 {
|
Chris@145
|
266 QString transformId = transform.getIdentifier();
|
Chris@145
|
267 if (m_tasks.find(transformId) != m_tasks.end()) return;
|
Chris@145
|
268
|
Chris@145
|
269 loadRDFDescription(transform);
|
Chris@145
|
270
|
Chris@145
|
271 Task task = UnknownTask;
|
Chris@145
|
272
|
Chris@145
|
273 QString pluginId = transform.getPluginIdentifier();
|
Chris@145
|
274 QString outputId = transform.getOutput();
|
Chris@145
|
275
|
Chris@145
|
276 const PluginRDFDescription &desc = m_rdfDescriptions[pluginId];
|
Chris@145
|
277
|
Chris@145
|
278 if (desc.haveDescription()) {
|
Chris@145
|
279
|
Chris@145
|
280 PluginRDFDescription::OutputDisposition disp =
|
Chris@145
|
281 desc.getOutputDisposition(outputId);
|
Chris@145
|
282
|
Chris@145
|
283 QString af = "http://purl.org/ontology/af/";
|
Chris@145
|
284
|
Chris@145
|
285 if (disp == PluginRDFDescription::OutputSparse) {
|
Chris@145
|
286
|
Chris@145
|
287 QString eventUri = desc.getOutputEventTypeURI(outputId);
|
Chris@145
|
288
|
Chris@145
|
289 //!!! todo: allow user to prod writer for task type
|
Chris@145
|
290
|
Chris@145
|
291 if (eventUri == af + "Note") {
|
Chris@145
|
292 task = NoteTask;
|
Chris@145
|
293 } else if (eventUri == af + "Beat") {
|
Chris@145
|
294 task = BeatTask;
|
Chris@145
|
295 } else if (eventUri == af + "ChordSegment") {
|
Chris@145
|
296 task = ChordTask;
|
Chris@145
|
297 } else if (eventUri == af + "KeyChange") {
|
Chris@145
|
298 task = KeyTask;
|
Chris@145
|
299 } else if (eventUri == af + "KeySegment") {
|
Chris@145
|
300 task = KeyTask;
|
Chris@145
|
301 } else if (eventUri == af + "Onset") {
|
Chris@145
|
302 task = OnsetTask;
|
Chris@145
|
303 } else if (eventUri == af + "NonTonalOnset") {
|
Chris@145
|
304 task = OnsetTask;
|
Chris@145
|
305 } else if (eventUri == af + "Segment") {
|
Chris@145
|
306 task = SegmentTask;
|
Chris@145
|
307 } else if (eventUri == af + "SpeechSegment") {
|
Chris@145
|
308 task = SegmentTask;
|
Chris@145
|
309 } else if (eventUri == af + "StructuralSegment") {
|
Chris@145
|
310 task = SegmentTask;
|
Chris@145
|
311 } else {
|
Chris@145
|
312 cerr << "WARNING: Unsupported event type URI <"
|
Chris@145
|
313 << eventUri << ">, proceeding with UnknownTask type"
|
Chris@145
|
314 << endl;
|
Chris@145
|
315 }
|
Chris@145
|
316
|
Chris@145
|
317 } else {
|
Chris@145
|
318
|
Chris@145
|
319 cerr << "WARNING: Cannot currently write dense or track-level outputs to JSON format (only sparse ones). Will proceed using UnknownTask type, but this probably isn't going to work" << endl;
|
Chris@145
|
320 }
|
Chris@145
|
321 }
|
Chris@145
|
322
|
Chris@145
|
323 m_tasks[transformId] = task;
|
Chris@145
|
324 }
|
Chris@145
|
325
|
Chris@145
|
326 QString
|
Chris@145
|
327 JAMSFeatureWriter::getTaskKey(Task task)
|
Chris@145
|
328 {
|
Chris@145
|
329 switch (task) {
|
Chris@145
|
330 case UnknownTask: return "unknown";
|
Chris@145
|
331 case BeatTask: return "beat";
|
Chris@145
|
332 case OnsetTask: return "onset";
|
Chris@145
|
333 case ChordTask: return "chord";
|
Chris@145
|
334 case SegmentTask: return "segment";
|
Chris@145
|
335 case KeyTask: return "key";
|
Chris@145
|
336 case NoteTask: return "note";
|
Chris@145
|
337 case MelodyTask: return "melody";
|
Chris@145
|
338 case PitchTask: return "pitch";
|
Chris@145
|
339 }
|
Chris@145
|
340 return "unknown";
|
Chris@145
|
341 }
|
Chris@165
|
342
|
Chris@165
|
343 QString
|
Chris@165
|
344 JAMSFeatureWriter::writeTransformToObjectContents(const Transform &t)
|
Chris@165
|
345 {
|
Chris@165
|
346 QString json;
|
Chris@165
|
347 QString stpl(" \"%1\": \"%2\",\n");
|
Chris@165
|
348 QString ntpl(" \"%1\": %2,\n");
|
Chris@165
|
349
|
Chris@165
|
350 json += stpl.arg("plugin_id").arg(t.getPluginIdentifier());
|
Chris@165
|
351 json += stpl.arg("output_id").arg(t.getOutput());
|
Chris@165
|
352
|
Chris@165
|
353 if (t.getSummaryType() != Transform::NoSummary) {
|
Chris@165
|
354 json += stpl.arg("summary_type")
|
Chris@165
|
355 .arg(Transform::summaryTypeToString(t.getSummaryType()));
|
Chris@165
|
356 }
|
Chris@165
|
357
|
Chris@165
|
358 if (t.getPluginVersion() != QString()) {
|
Chris@165
|
359 json += stpl.arg("plugin_version").arg(t.getPluginVersion());
|
Chris@165
|
360 }
|
Chris@165
|
361
|
Chris@165
|
362 if (t.getProgram() != QString()) {
|
Chris@165
|
363 json += stpl.arg("program").arg(t.getProgram());
|
Chris@165
|
364 }
|
Chris@165
|
365
|
Chris@165
|
366 if (t.getStepSize() != 0) {
|
Chris@165
|
367 json += ntpl.arg("step_size").arg(t.getStepSize());
|
Chris@165
|
368 }
|
Chris@165
|
369
|
Chris@165
|
370 if (t.getBlockSize() != 0) {
|
Chris@165
|
371 json += ntpl.arg("block_size").arg(t.getBlockSize());
|
Chris@165
|
372 }
|
Chris@165
|
373
|
Chris@165
|
374 if (t.getWindowType() != HanningWindow) {
|
Chris@165
|
375 json += stpl.arg("window_type")
|
Chris@165
|
376 .arg(Window<float>::getNameForType(t.getWindowType()).c_str());
|
Chris@165
|
377 }
|
Chris@165
|
378
|
Chris@165
|
379 if (t.getStartTime() != RealTime::zeroTime) {
|
Chris@165
|
380 json += ntpl.arg("start").arg(t.getStartTime().toDouble());
|
Chris@165
|
381 }
|
Chris@165
|
382
|
Chris@165
|
383 if (t.getDuration() != RealTime::zeroTime) {
|
Chris@165
|
384 json += ntpl.arg("duration").arg(t.getDuration().toDouble());
|
Chris@165
|
385 }
|
Chris@165
|
386
|
Chris@165
|
387 if (t.getSampleRate() != 0) {
|
Chris@165
|
388 json += ntpl.arg("sample_rate").arg(t.getSampleRate());
|
Chris@165
|
389 }
|
Chris@165
|
390
|
Chris@165
|
391 if (!t.getParameters().empty()) {
|
Chris@165
|
392 json += QString(" \"parameters\": {\n");
|
Chris@165
|
393 Transform::ParameterMap parameters = t.getParameters();
|
Chris@165
|
394 for (Transform::ParameterMap::const_iterator i = parameters.begin();
|
Chris@165
|
395 i != parameters.end(); ++i) {
|
Chris@165
|
396 QString name = i->first;
|
Chris@165
|
397 float value = i->second;
|
Chris@165
|
398 json += QString(" \"%1\": %2\n").arg(name).arg(value);
|
Chris@165
|
399 }
|
Chris@165
|
400 json += QString(" },\n");
|
Chris@165
|
401 }
|
Chris@165
|
402
|
Chris@165
|
403 // no trailing comma on final property:
|
Chris@165
|
404 json += QString(" \"transform_id\": \"%1\"\n").arg(t.getIdentifier());
|
Chris@165
|
405
|
Chris@165
|
406 return json;
|
Chris@165
|
407 }
|
Chris@165
|
408
|