Audio Reconstruction on AudioCaps and Song Describer Dataset (60 clips)
4.45MOSGround Truth
Evaluation results
| Method | Links | |
|---|---|---|
| Ground TruthLatent Rate (Hz)=–2026.06 | 4.45 | |
| STAR-VAELatent Rate (Hz)=21.52026.06 | 4.32 | |
| ϵar-VAELatent Rate (Hz)=432026.06 | 4.21 | |
| SAOLatent Rate (Hz)=21.52026.06 | 4.05 |