Action Recognition on Kinetics-Sounds (test)
91.44Top-1 AccuracyIMD-B*
Evaluation Results
| Method | Links | |
|---|---|---|
| IMD-B*P-train=IN-21K, intra-class cross-modality augmentation=without2022.03 | 91.44 | |
| IMD-BP-train=IN-21K, intra-class cross-modality augmentation=with2022.03 | 90.48 | |
| AVSlowFast, R101P-train=IN-1K2022.03 | 85 | |
| MBT (AV)P-train=IN-21K2022.03 | 85 | |
| XKDModality=Audio + Video, Fusion Strategy=Late fusion2022.11 | 81.2 | |
| XKD-MASModality=Audio + Video, Fusion Strategy=Late fusion2022.11 | 78.8 | |
| XKD-MATSModality=Audio + Video, Fusion Strategy=Late fusion2022.11 | 78.3 | |
| SlowFast R101P-train=IN-1K2022.03 | 77.9 | |
| L_ae (AV-MAE)Modality=Audio + Video, Fusion Strategy=Late fusion2022.11 | 75.7 | |
| L3-NetP-train=IN-1K2022.03 | 74 | |
| CrissCrossModality=Audio + Video, Fusion Strategy=Late fusion2022.11 | 66.7 |