Single-target Sound Extraction on FSD Kaggle 2018 and TAU Urban Acoustic Scenes 2019 (test)
9.43SI-SNRiWaveformer
Evaluation results
| Method | Links | |
|---|---|---|
| WaveformerE=512, D=256, Model size=3.88M, RTF=0.942022.11 | 9.43 | |
| WaveformerE=256, D=256, Model size=1.69M, RTF=0.752022.11 | 9.4 | |
| WaveformerE=512, D=128, Model size=3.29M, RTF=0.882022.11 | 9.26 | |
| WaveformerE=256, D=128, Model size=1.10M, RTF=0.662022.11 | 9.02 | |
| ReSepformerModel size=13.24M, RTF=1.602022.11 | 7.26 | |
| Conv-TasNetModel size=4.57M, RTF=1.342022.11 | 6.14 |