Continual Audio-Visual Sound Separation on VGGSound 13 (test)
5.57SDRUpper Bound
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Upper BoundBackbone=iQuery2024.11 | 5.57 | 9.19 | 13.24 | |
| ContAV-SepContinual learning strategy=Continual Audio-Visual Sound Separation2024.11 | 4.9 | 9.25 | 11.73 | |
| iQuery + LwFContinual learning strategy=LwF, Memory usage=20 samples per class, Backbone=iQuery2024.11 | 4.71 | 8.89 | 11.7 | |
| iQuery + AV-CILContinual learning strategy=AV-CIL, Memory usage=20 samples per class, Backbone=iQuery2024.11 | 4.69 | 8.61 | 11.6 | |
| iQuery + PLOPContinual learning strategy=PLOP, Memory usage=20 samples per class, Backbone=iQuery2024.11 | 4.56 | 8.32 | 12.34 | |
| iQuery + Fine-tuningContinual learning strategy=Fine-tuning, Backbone=iQuery2024.11 | 3.69 | 7.23 | 12.84 |