Voice Activity Detection on AVA-Speech (test)
94.3AUC-ROCSG-VAD
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SG-VADParams=7.8k2022.10 | 94.3 | — | — | — | — | |
| Braun et al.Params=1773k, Training Dataset=DNS2022.10 | 92.4 | — | — | — | — | |
| NAS-VADParams=151k, Training Dataset=CV, AS2022.10 | 90.5 | — | — | — | — | |
| ResectNet 1.0xParams=11.1k, Training Dataset=DNS2022.10 | 90 | — | — | — | — | |
| ResectNet 0.5xParams=4.5k, Training Dataset=DNS2022.10 | 88.6 | — | — | — | — | |
| MarbleNetParams=88k, Training Dataset=GSCV2, FS2K2022.10 | 85.8 | — | — | — | — | |
| ADA-VADTraining Dataset=TIMIT2022.10 | 85.3 | — | — | — | — | |
| CNN-BILSTMbestFalse Positive Rate (FPR)=0.315, Frame duration=10ms, Parameters=516k, Evaluation protocol=Mean of outer fold cross-validation2021.03 | — | 99.2 | 96.1 | 95 | 96.8 | |
| CNN-BILSTMsmallFalse Positive Rate (FPR)=0.315, Frame duration=10ms, Parameters=192k, Evaluation protocol=Mean of outer fold cross-validation2021.03 | — | 99.2 | 96 | 94.8 | 96.7 | |
| resnet960False Positive Rate (FPR)=0.315, Frame duration=10ms, Backbone=ResNet-50, Parameters=30M2021.03 | — | 99.2 | 94.4 | 78.7 | 91.7 | |
| RTCvadFalse Positive Rate (FPR)=0.315, Frame duration=10ms2021.03 | — | 78.6 | 70.6 | 73.3 | 72.2 | |
| tiny320False Positive Rate (FPR)=0.315, Frame duration=10ms, Parameters=<1M2021.03 | — | 96.5 | 82.6 | 62.3 | 81 |