Spherical Active Speaker Localization on EasyCom (full)
8.81Mean E1 ErrorEgocentric Deep Multi-Channel Audio-Visual Network
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Egocentric Deep Multi-Channel Audio-Visual NetworkInput Modality=Audio-Visual, Audio Feature Type=Spectrogram (spec)2022.01 | 8.81 | 9.63 | 6.21 | 6.89 | |
| Egocentric Deep Multi-Channel Audio-Visual NetworkInput Modality=Audio-Visual, Audio Feature Type=Correlation (cor)2022.01 | 16.77 | 12.63 | 6.56 | 8.77 | |
| AV-rawaudioInput Modality=Audio-Visual, Audio Feature Type=raw audio input2022.01 | 40.14 | 10.55 | 140.75 | 19.58 | |
| DOA+imageInput Modality=Audio-Visual (DOA maps + visual)2022.01 | 66.81 | 7.89 | 36.48 | 8.97 | |
| DOAInput Modality=Audio-only (DOA maps)2022.01 | 129.82 | 18.26 | 46.45 | 21.5 |