Audio-visual event classification on AudioSet 20K (test)
42.4mAP (Audio)EquiAV
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| EquiAVEvaluation Protocol=Full-model finetuning (reported), Params=170M, Pre-train data=IN+AS2026.06 | 42.4 | 25.7 | 46.6 | |
| MAViLEvaluation Protocol=Full-model finetuning (reported), Params=170M, Pre-train data=IN+AS2026.06 | 41.8 | 24.8 | 44.9 | |
| MJEPA ViT-g + data scalingEvaluation Protocol=Frozen evaluation, Params=1B, Pre-train data=AS+VM2M, Backbone=ViT-g2026.06 | 40.97 | 29.82 | 45.44 | |
| MJEPA ViT-L + data scalingEvaluation Protocol=Frozen evaluation, Params=300M, Pre-train data=AS+VM2M, Backbone=ViT-L2026.06 | 40 | 29.63 | 45.31 | |
| MJEPA ViT-LEvaluation Protocol=Frozen evaluation, Params=300M, Pre-train data=AS, Backbone=ViT-L2026.06 | 38.89 | 25.38 | 42.9 | |
| CAV-MAEEvaluation Protocol=Full-model finetuning (reported), Params=170M, Pre-train data=IN+AS2026.06 | 37.7 | 19.8 | 42 | |
| EquiAVEvaluation Protocol=Frozen evaluation, Params=170M, Pre-train data=IN+AS2026.06 | 34.25 | 18.6 | 38.6 | |
| MAViLEvaluation Protocol=Frozen evaluation, Params=170M, Pre-train data=IN+AS2026.06 | 30 | — | — | |
| CAV-MAE SyncEvaluation Protocol=Frozen evaluation, Params=170M, Pre-train data=IN+AS2026.06 | 21.66 | 16.2 | 28.5 | |
| CAV-MAEEvaluation Protocol=Frozen evaluation, Params=170M, Pre-train data=IN+AS2026.06 | 19.38 | 18.14 | 34.59 |