Spaces:
Running on Zero
Running on Zero
File size: 1,682 Bytes
127b976 | 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 |
from __future__ import annotations
from typing import Dict, List
import numpy as np
import torch
import torchaudio
def load_audio_mono(path: str, target_sr: int = 16000):
wav, sr = torchaudio.load(path)
wav = wav.mean(dim=0)
if sr != target_sr:
wav = torchaudio.functional.resample(wav, sr, target_sr)
sr = target_sr
return wav, sr
def audio_quality_report(path: str, target_sr: int = 16000) -> Dict:
wav, sr = load_audio_mono(path, target_sr)
x = wav.detach().cpu().numpy().astype(float)
if len(x) == 0:
return {"warnings": ["Audio appears empty."], "metrics": {"duration_sec": 0.0}}
duration = len(x) / sr
rms = float(np.sqrt(np.mean(x ** 2)) + 1e-12)
peak = float(np.max(np.abs(x)) + 1e-12)
silence_ratio = float(np.mean(np.abs(x) < 1e-4))
clipping_ratio = float(np.mean(np.abs(x) > 0.98))
warnings: List[str] = []
if duration < 1.0:
warnings.append("Audio is very short; prediction may be unreliable.")
if rms < 0.005:
warnings.append("Audio volume is very low.")
if clipping_ratio > 0.01:
warnings.append("Audio may be clipped or distorted.")
if silence_ratio > 0.70:
warnings.append("Audio contains a lot of silence.")
if peak < 0.02:
warnings.append("Audio peak level is very low.")
return {
"warnings": warnings,
"metrics": {
"sample_rate": sr,
"duration_sec": duration,
"rms": rms,
"peak": peak,
"silence_ratio": silence_ratio,
"clipping_ratio": clipping_ratio,
},
}
|