Spaces:
Sleeping
Sleeping
File size: 11,498 Bytes
b77cba7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 |
#!/usr/bin/env python3
"""
Integrated VAD + Speaker Diarization Pipeline
Real-time processing with optimized performance
"""
import torch
import numpy as np
from typing import List, Dict, Optional, Tuple, Union
import time
from pathlib import Path
import json
from .vad import SileroVAD
from .diarization import SpeakerDiarization
class VADDiarizationPipeline:
"""
Integrated pipeline combining VAD and speaker diarization.
Features:
- Two-stage processing: VAD first, then diarization
- Optimized for real-time performance
- Configurable parameters
- Comprehensive output format
"""
def __init__(
self,
vad_threshold: float = 0.5,
use_auth_token: Optional[str] = None,
token: Optional[str] = None,
device: Optional[str] = None,
num_speakers: Optional[int] = None,
min_speakers: Optional[int] = None,
max_speakers: Optional[int] = None,
use_onnx_vad: bool = False
):
"""
Initialize the integrated pipeline.
Args:
vad_threshold: VAD sensitivity threshold
use_auth_token: (Deprecated) Hugging Face token for diarization
token: Hugging Face token for diarization (new parameter name)
device: Device to use ('cuda' or 'cpu')
num_speakers: Fixed number of speakers
min_speakers: Minimum number of speakers
max_speakers: Maximum number of speakers
use_onnx_vad: Use ONNX for VAD (faster)
"""
print("\n" + "="*60)
print("INITIALIZING VAD + DIARIZATION PIPELINE")
print("="*60)
# Handle both old and new parameter names
auth_token = token or use_auth_token
# Initialize VAD
print("\n[1/2] Loading Voice Activity Detection...")
self.vad = SileroVAD(
threshold=vad_threshold,
use_onnx=use_onnx_vad
)
# Initialize Diarization
print("\n[2/2] Loading Speaker Diarization...")
self.diarization = SpeakerDiarization(
token=auth_token,
device=device,
num_speakers=num_speakers,
min_speakers=min_speakers,
max_speakers=max_speakers
)
print("\n" + "="*60)
print("✅ PIPELINE READY")
print("="*60 + "\n")
def process_file(
self,
audio_path: str,
num_speakers: Optional[int] = None,
return_vad: bool = True,
return_stats: bool = True
) -> Dict:
"""
Process an audio file through the complete pipeline.
Args:
audio_path: Path to audio file
num_speakers: Number of speakers (if known)
return_vad: Include VAD segments in output
return_stats: Include statistics in output
Returns:
Dict with results and metadata
"""
print(f"\n📁 Processing: {audio_path}")
print("-" * 60)
total_start = time.time()
# Stage 1: VAD
print("Stage 1: Voice Activity Detection...")
vad_start = time.time()
vad_segments, vad_time = self.vad.process_file(audio_path)
vad_duration = (time.time() - vad_start) * 1000
print(f" ✓ Found {len(vad_segments)} speech segments")
print(f" ✓ Processing time: {vad_duration:.2f}ms")
# Stage 2: Diarization
print("\nStage 2: Speaker Diarization...")
diar_start = time.time()
speaker_segments, diar_time, diar_metadata = self.diarization.process_file(
audio_path,
num_speakers=num_speakers
)
diar_duration = (time.time() - diar_start) * 1000
print(f" ✓ Identified {diar_metadata['num_speakers']} speakers")
print(f" ✓ Found {diar_metadata['num_segments']} speaker segments")
print(f" ✓ Processing time: {diar_duration:.2f}ms")
# Calculate total time
total_duration = (time.time() - total_start) * 1000
print(f"\n⏱️ Total processing time: {total_duration:.2f}ms")
print("-" * 60)
# Build result
result = {
'audio_path': audio_path,
'speaker_segments': speaker_segments,
'processing_time': {
'vad_ms': vad_duration,
'diarization_ms': diar_duration,
'total_ms': total_duration
},
'metadata': diar_metadata
}
if return_vad:
result['vad_segments'] = vad_segments
if return_stats:
result['speaker_statistics'] = self.diarization.get_speaker_statistics(
speaker_segments
)
return result
def process_batch(
self,
audio_paths: List[str],
**kwargs
) -> List[Dict]:
"""
Process multiple audio files.
Args:
audio_paths: List of audio file paths
**kwargs: Additional arguments for process_file
Returns:
List of results
"""
results = []
print(f"\n📦 Batch processing {len(audio_paths)} files...")
print("="*60)
for i, path in enumerate(audio_paths, 1):
print(f"\n[{i}/{len(audio_paths)}]")
result = self.process_file(path, **kwargs)
results.append(result)
print("\n" + "="*60)
print(f"✅ Batch processing complete ({len(results)} files)")
print("="*60 + "\n")
return results
def format_output(self, result: Dict, format: str = 'text') -> str:
"""
Format pipeline output.
Args:
result: Result from process_file
format: Output format ('text', 'json', 'rttm')
Returns:
Formatted string
"""
if format == 'json':
return json.dumps(result, indent=2)
elif format == 'rttm':
# RTTM format for NIST evaluation
lines = []
for seg in result['speaker_segments']:
# RTTM format: SPEAKER file 1 start duration <NA> <NA> speaker <NA> <NA>
line = f"SPEAKER {Path(result['audio_path']).stem} 1 {seg['start']:.3f} {seg['duration']:.3f} <NA> <NA> {seg['speaker']} <NA> <NA>"
lines.append(line)
return "\n".join(lines)
else: # text
lines = []
lines.append("="*60)
lines.append("VAD + SPEAKER DIARIZATION RESULTS")
lines.append("="*60)
lines.append(f"\nFile: {result['audio_path']}")
# Metadata
lines.append(f"\nMetadata:")
lines.append(f" Speakers: {result['metadata']['num_speakers']}")
lines.append(f" Segments: {result['metadata']['num_segments']}")
lines.append(f" Total speech: {result['metadata']['total_speech_time']:.2f}s")
# Processing time
lines.append(f"\nProcessing Time:")
lines.append(f" VAD: {result['processing_time']['vad_ms']:.2f}ms")
lines.append(f" Diarization: {result['processing_time']['diarization_ms']:.2f}ms")
lines.append(f" Total: {result['processing_time']['total_ms']:.2f}ms")
# Speaker statistics
if 'speaker_statistics' in result:
lines.append(f"\nSpeaker Statistics:")
for speaker, stats in result['speaker_statistics'].items():
lines.append(f" {speaker}:")
lines.append(f" Total time: {stats['total_time']:.2f}s")
lines.append(f" Segments: {stats['num_segments']}")
lines.append(f" Avg duration: {stats['avg_segment_duration']:.2f}s")
# Timeline
lines.append(f"\nSpeaker Timeline:")
lines.append("-"*60)
for seg in result['speaker_segments']:
lines.append(f"{seg['start']:7.2f}s - {seg['end']:7.2f}s: {seg['speaker']}")
lines.append("="*60)
return "\n".join(lines)
def save_results(
self,
result: Dict,
output_path: str,
format: str = 'json'
):
"""
Save results to file.
Args:
result: Result from process_file
output_path: Output file path
format: Output format ('json', 'rttm', 'text')
"""
output = self.format_output(result, format=format)
with open(output_path, 'w') as f:
f.write(output)
print(f"✓ Results saved to: {output_path}")
def benchmark(
self,
test_audio_path: Optional[str] = None,
duration_seconds: float = 10.0
) -> Dict:
"""
Benchmark pipeline performance.
Args:
test_audio_path: Path to test audio (optional)
duration_seconds: Duration for synthetic test
Returns:
Benchmark metrics
"""
print("\n" + "="*60)
print("PIPELINE BENCHMARK")
print("="*60)
# VAD benchmark
print("\n[1/2] Benchmarking VAD...")
vad_metrics = self.vad.benchmark_latency(duration_seconds)
print(f" Latency: {vad_metrics['latency_per_second_ms']:.2f}ms per second")
print(f" Real-time factor: {vad_metrics['real_time_factor']:.4f}x")
if vad_metrics['latency_per_second_ms'] < 100:
print(" ✅ VAD latency target achieved (<100ms)")
else:
print(" ⚠️ VAD latency above target")
# Full pipeline benchmark (if test audio provided)
if test_audio_path:
print("\n[2/2] Benchmarking full pipeline...")
result = self.process_file(test_audio_path, return_stats=False)
print(f" Total time: {result['processing_time']['total_ms']:.2f}ms")
print("\n" + "="*60)
return {
'vad_metrics': vad_metrics,
'pipeline_metrics': result['processing_time'] if test_audio_path else None
}
def demo():
"""Demo the integrated pipeline."""
print("\n" + "="*60)
print("INTEGRATED PIPELINE DEMO")
print("="*60)
import os
# Check for HF token
token = os.environ.get('HF_TOKEN')
if not token:
print("\n⚠️ No HF_TOKEN found in environment")
print("Set it with: export HF_TOKEN='your_token_here'")
print("\nFor now, will demo VAD only...")
# VAD-only demo
vad = SileroVAD()
metrics = vad.benchmark_latency()
print(f"\n✅ VAD latency: {metrics['latency_per_second_ms']:.2f}ms per second")
return
try:
# Initialize pipeline
pipeline = VADDiarizationPipeline(
use_auth_token=token,
vad_threshold=0.5
)
# Benchmark
pipeline.benchmark()
print("\n✅ Pipeline demo complete!")
except Exception as e:
print(f"\n❌ Error: {e}")
print("\n" + "="*60)
if __name__ == "__main__":
demo()
|