Material Classification on Greatest Hits
61AccuracyMMAudio-LABEL (Joint Heads (finetune))
Evaluation Results
| Method | Links | |
|---|---|---|
| MMAudio-LABEL (Joint Heads (finetune))Training=Finetune, Input=visual, Output=label + audio, Note=finetuned from small-16k checkpoint2026.05 | 61 | |
| MMAudio-LABEL (Joint Heads)Training=Scratch, Input=visual, Output=label + audio2026.05 | 51.9 | |
| MMAudio-LABEL (Parallel Heads)Training=Scratch, Input=visual, Output=label + audio2026.05 | 43.9 | |
| VGGish ClassifierTraining=Scratch, Input=audio, Output=label2026.05 | 40.6 | |
| MMAudio-LABEL (Event Head Only)Training=Scratch, Input=visual, Output=label2026.05 | 39 |