Multi-label Sound Event Classification on FSD50K
65.8mAPMJEPA ViT-g + data scaling
Evaluation results
| Method | Links | |
|---|---|---|
| MJEPA ViT-g + data scalingEvaluation protocol=Frozen feature probing, Params=1B, Modality=A+V, Pre-train data=AS+VM2M2026.06 | 65.8 | |
| MJEPA ViT-L + data scalingEvaluation protocol=Frozen feature probing, Params=300M, Modality=A+V, Pre-train data=AS+VM2M2026.06 | 65.5 | |
| MJEPA ViT-LEvaluation protocol=Frozen feature probing, Params=300M, Modality=A+V, Pre-train data=AS2026.06 | 63.9 | |
| EquiAVEvaluation protocol=Full-model finetuning, Params=170M, Modality=A+V, Pre-train data=AS2026.06 | 62.6 | |
| XKDEvaluation protocol=Full-model finetuning, Params=170M, Modality=A+V, Pre-train data=AS2026.06 | 58.5 | |
| SSLAMEvaluation protocol=Frozen feature probing, Params=88M, Modality=A, Pre-train data=AS2026.06 | 57.9 | |
| EquiAVEvaluation protocol=Frozen feature probing, Params=170M, Modality=A+V, Pre-train data=AS2026.06 | 57.9 | |
| SPEAREvaluation protocol=Frozen feature probing, Params=600M, Modality=A, Pre-train data=197k hrs mix2026.06 | 57.1 | |
| Dasheng-1.2BEvaluation protocol=Frozen feature probing, Params=1.2B, Modality=A, Pre-train data=272k hrs mix2026.06 | 56.9 | |
| CAV-MAE SyncEvaluation protocol=Frozen feature probing, Params=170M, Modality=A+V, Pre-train data=AS2026.06 | 55.5 | |
| XKDEvaluation protocol=Frozen feature probing, Params=170M, Modality=A+V, Pre-train data=AS2026.06 | 51.5 | |
| CAV-MAEEvaluation protocol=Frozen feature probing, Params=170M, Modality=A+V, Pre-train data=IN+AS2026.06 | 46.1 | |
| AJEPAEvaluation protocol=Frozen feature probing, Params=85M, Modality=A, Pre-train data=AS2026.06 | 43.9 |