Audio-Visual Speech Separation on LRS3 (test)
18.9SDRiDolphin
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Dolphin2025.09 | 18.9 | 3.36 | — | 18.8 | |
| IIANet2023.08 | 18.5 | 3.28 | — | 18.3 | |
| IIANetresults_source=reported directly from original paper2025.09 | 18.5 | 3.28 | — | 18.3 | |
| AV-Mossformer2results_source=reproduced using official code2025.09 | 18.1 | 3.28 | — | 17.7 | |
| IIANet-fast2023.08 | 17.9 | 3.25 | — | 17.8 | |
| RTFS-Netresults_source=reported directly from original paper2025.09 | 17.6 | 3.25 | — | 17.5 | |
| CTCNet2023.08 | 17.5 | 3.24 | — | 17.4 | |
| CTCNetresults_source=reported directly from original paper2025.09 | 17.5 | 3.24 | — | 17.4 | |
| Swift-Netresults_source=reproduced using official code2025.09 | 16.4 | 3.11 | — | 15.8 | |
| AVLIT-82023.08 | 13.6 | 2.78 | — | 13.5 | |
| AVLiT-8results_source=reported directly from original paper2025.09 | 13.6 | 2.78 | — | 13.5 | |
| CaffNet-CPhase estimation=Ground-truth (GT)2021.03 | 12.31 | 0.91 | 0.86 | — | |
| AVConvTasNet2023.08 | 11.7 | 2.58 | — | 11.2 | |
| AV-ConvTasNetresults_source=reported directly from original paper2025.09 | 11.7 | 2.58 | — | 11.2 | |
| V-ConvPhase estimation=Ground-truth (GT)2021.03 | 11.23 | 1.08 | 0.86 | — | |
| Visualvoice2023.08 | 10.3 | 2.13 | — | 9.9 | |
| Visualvoiceresults_source=reported directly from original paper2025.09 | 10.3 | 2.13 | — | 9.9 | |
| CaffNetPhase estimation=Ground-truth (GT)2021.03 | 10.22 | 1.06 | 0.86 | — | |
| CaffNet-CPhase estimation=Predicted phase (PR)2021.03 | 9.78 | 0.71 | 0.85 | — | |
| CaffNet-CPhase estimation=Mixture phase (MX)2021.03 | 7.92 | 0.55 | 0.83 | — | |
| LWTNetPhase estimation=Ground-truth (GT)2021.03 | 7.71 | 0.82 | 0.84 | — | |
| V-ConvPhase estimation=Mixture phase (MX)2021.03 | 7 | 0.61 | 0.83 | — | |
| CaffNetPhase estimation=Mixture phase (MX)2021.03 | 6.46 | 0.6 | 0.84 | — | |
| LWTNetPhase estimation=Predicted phase (PR)2021.03 | 4.83 | 0.49 | 0.82 | — | |
| LWTNetPhase estimation=Mixture phase (MX)2021.03 | 4.44 | 0.45 | 0.82 | — | |
| V-ConvPhase estimation=Griffin-Lim (GL)2021.03 | -1.37 | 0.55 | 0.82 | — | |
| CaffNet-CPhase estimation=Griffin-Lim (GL)2021.03 | -1.38 | 0.49 | 0.82 | — | |
| CaffNetPhase estimation=Griffin-Lim (GL)2021.03 | -2.64 | 0.49 | 0.82 | — | |
| LWTNetPhase estimation=Griffin-Lim (GL)2021.03 | -3.93 | 0.35 | 0.8 | — | |
| CaffNet-C2023.08 | — | — | — | 9.8 | |
| LWTNet2023.08 | — | — | — | 4.8 |