| Benchmark | Model | Rank | Results |
|---|---|---|---|
| action-recognition-in-videos-on-ava-v21 | ACAR-Net, SlowFast R-101 (Kinetics-400 pretraining) | #1 | mAP (Val): 30.0 |
| action-recognition-on-ava-v2-2 | ACAR-Net, SlowFast R-101 (Kinetics-700 pretraining) | #22 | mAP: 31.72 |
| spatio-temporal-action-localization-on-ava | ACAR (multi-scale, ensemble) | #3 | val mAP: 40.49test mAP: 39.62 |
| spatio-temporal-action-localization-on-ava | ACAR (multi-scale, R-101, 8 × 8) | #4 | val mAP: 36.36 |