Speech Restoration on REVERB Sim
2.11CDTF-GridNet
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| TF-GridNettraining_condition=in-domain (trained on REVERB), model_type=discriminative baseline2025.09 | 2.11 | 0.216 | 14.458 | 2.825 | 5.063 | 3.325 | 3.135 | 3.31 | 3.435 | 3.945 | |
| TF-Locoformertraining_condition=in-domain (trained on REVERB), model_type=discriminative baseline2025.09 | 2.323 | 0.235 | 14.102 | 2.642 | 4.972 | 3.202 | 3.063 | 3.056 | 3.366 | 3.92 | |
| MP-SENettraining_condition=in-domain (trained on REVERB), model_type=discriminative baseline2025.09 | 2.692 | 0.278 | 13.622 | 2.811 | 4.96 | 3.225 | 3.19 | 3.5 | 3.45 | 4.035 | |
| TF-Restormertraining_condition=OOD restoration, setup=single unified model2025.09 | 2.88 | 0.489 | 8.139 | 2.825 | 8.004 | 4.13 | 3.335 | 4.86 | 3.585 | 4.115 | |
| VoiceFixertraining_condition=OOD restoration, setup=matched conditions2025.09 | 3.708 | 0.946 | 7.992 | 1.278 | 4.938 | 2.858 | 3.088 | 4.072 | 3.388 | 3.927 | |
| Inputcondition=unprocessed2025.09 | 3.975 | 0.574 | 3.617 | 1.503 | 3.687 | 1.455 | 1.365 | 1.255 | 1.75 | 1.39 | |
| UNIVERSE++training_condition=OOD restoration, setup=matched conditions2025.09 | 5.213 | 0.901 | 5.731 | 1.22 | 5.124 | 2.37 | 2.835 | 4.145 | 3.15 | 3.835 |