docs: frame F1 97.4% vs FunASR
Browse files
README.md
CHANGED
|
@@ -37,8 +37,14 @@ waveform → [Preprocessor fp32/CPU] → features [1,T,400]
|
|
| 37 |
`max_start_silence_time` (3000 ms), `max_single_segment_time` (60 s),
|
| 38 |
`sil_to_speech_time_thres` (150 ms). See `vad_config.json`.
|
| 39 |
|
| 40 |
-
|
| 41 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
|
| 43 |
## License
|
| 44 |
|
|
|
|
| 37 |
`max_start_silence_time` (3000 ms), `max_single_segment_time` (60 s),
|
| 38 |
`sil_to_speech_time_thres` (150 ms). See `vad_config.json`.
|
| 39 |
|
| 40 |
+
## Benchmark — fidelity vs FunASR (FLEURS zh, n=50)
|
| 41 |
+
|
| 42 |
+
| Metric | Value |
|
| 43 |
+
|--------|-------|
|
| 44 |
+
| **Frame F1** | **97.4%** (P 100.0% / R 94.8%) |
|
| 45 |
+
| Median RTFx | 1209x |
|
| 46 |
+
|
| 47 |
+
Parity: preprocessor matches `WavFrontendOnline` max|Δ|≈3e-5; FSMN scorer max|Δ| 0.0016. Boundaries match FunASR within ~50 ms.
|
| 48 |
|
| 49 |
## License
|
| 50 |
|