Multi-label Sound Event Classification on AS 20K
42.4mAPEquiAV
Evaluation results
| Method | Links | |
|---|---|---|
| EquiAVEvaluation protocol=Full-model finetuning, Params=170M, Modality=A+V, Pre-train data=AS2026.06 | 42.4 | |
| MAViLEvaluation protocol=Full-model finetuning, Params=170M, Modality=A+V, Pre-train data=IN+AS2026.06 | 41.8 | |
| MJEPA ViT-g + data scalingEvaluation protocol=Frozen feature probing, Params=1B, Modality=A+V, Pre-train data=AS+VM2M2026.06 | 40.9 | |
| SSLAMEvaluation protocol=Full-model finetuning, Params=88M, Modality=A, Pre-train data=AS2026.06 | 40.9 | |
| MJEPA ViT-L + data scalingEvaluation protocol=Frozen feature probing, Params=300M, Modality=A+V, Pre-train data=AS+VM2M2026.06 | 40 | |
| SPEAREvaluation protocol=Full-model finetuning, Params=600M, Modality=A, Pre-train data=197k hrs mix2026.06 | 39.4 | |
| MJEPA ViT-LEvaluation protocol=Frozen feature probing, Params=300M, Modality=A+V, Pre-train data=AS2026.06 | 38.9 | |
| AJEPAEvaluation protocol=Full-model finetuning, Params=85M, Modality=A, Pre-train data=AS2026.06 | 38.4 | |
| CAV-MAEEvaluation protocol=Full-model finetuning, Params=170M, Modality=A+V, Pre-train data=IN+AS2026.06 | 37.7 | |
| EquiAVEvaluation protocol=Frozen feature probing, Params=170M, Modality=A+V, Pre-train data=AS2026.06 | 34.3 | |
| Dasheng-1.2BEvaluation protocol=Frozen feature probing, Params=1.2B, Modality=A, Pre-train data=272k hrs mix2026.06 | 31.6 | |
| SSLAMEvaluation protocol=Frozen feature probing, Params=88M, Modality=A, Pre-train data=AS2026.06 | 31.4 | |
| MAViLEvaluation protocol=Frozen feature probing, Params=170M, Modality=A+V, Pre-train data=IN+AS2026.06 | 30 | |
| CAV-MAE SyncEvaluation protocol=Frozen feature probing, Params=170M, Modality=A+V, Pre-train data=AS2026.06 | 21.7 | |
| CAV-MAEEvaluation protocol=Frozen feature probing, Params=170M, Modality=A+V, Pre-train data=IN+AS2026.06 | 19.4 | |
| AJEPAEvaluation protocol=Frozen feature probing, Params=85M, Modality=A, Pre-train data=AS2026.06 | 18 | |
| SPEAREvaluation protocol=Frozen feature probing, Params=600M, Modality=A, Pre-train data=197k hrs mix2026.06 | 11.9 |