Image-to-Audio Generation on BPSD
68.83Onset F1 (50ms)DAC Reconstruction
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| DAC ReconstructionType=Upper-bound2025.05 | 68.83 | 82.39 | 87.47 | 0.05 | |
| Multi-stageTraining Tasks/Data=OMR + Image-to-Audio + MIDI-to-Audio2025.05 | 50.91 | 70.4 | 79.96 | 0.137 | |
| Direct I2ATraining Tasks/Data=OMR + Image-to-Audio2025.05 | 48.67 | 64.3 | 74.01 | 0.098 | |
| Direct I2ATraining Tasks/Data=OMR + Image-to-Audio + MIDI-to-Audio2025.05 | 48.36 | 64.63 | 74.92 | 0.081 | |
| Multi-stage: Zeus → VirtuosoNet → MSDMethod Components=Zeus, VirtuosoNet, MSD2025.05 | 45.52 | 59.35 | 69.36 | 0.33 | |
| Direct I2ATraining Tasks/Data=YTSV-P (Image-to-Audio Only)2025.05 | 23.49 | 34.51 | 44.15 | 0.422 |