Action Localization on AVA 2.2
34.4mAP (center)MeMViT-24, 32x3, 1312^2
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MeMViT-24, 32x3, 1312^2Pre-train=K700, GFLOPs=620, Param (M)=52.6, Resolution=1312x13122022.01 | 34.4 | 35.4 | |
| MeMViT-24, 32x3Pre-train=K700, GFLOPs=211.7, Param (M)=52.62022.01 | 33.5 | 34.4 | |
| MeMViT-24, 32x3, 1312^2Pre-train=K600, GFLOPs=620, Param (M)=52.6, Resolution=1312x13122022.01 | 32.8 | 33.6 | |
| AIAPre-train=K7002022.01 | 32.3 | — | |
| MViT-24, 32x3Pre-train=K700, GFLOPs=204.4, Param (M)=51.32022.01 | 31.8 | 32.5 | |
| MeMViT-24, 32x3Pre-train=K600, GFLOPs=211.7, Param (M)=52.62022.01 | 31.5 | 32.3 | |
| ACAR 8x8, R101-NLPre-train=K600, GFLOPs=293.2, Param (M)=118.4+2022.01 | 31.4 | — | |
| Object TransformerPre-train=K600, GFLOPs=243.8, Param (M)=86.22022.01 | 31 | — | |
| MViT-24, 32x3Pre-train=K600, GFLOPs=204.4, Param (M)=51.32022.01 | 29.4 | 30.1 | |
| MViTv1-B-24, 32x3Pre-train=K600, GFLOPs=236, Param (M)=52.92022.01 | 28.7 | — | |
| MeMViT-16, 16x4Pre-train=K400, GFLOPs=58.7, Param (M)=35.42022.01 | 28.5 | 29.3 | |
| WOO, SFR101Pre-train=K600, GFLOPs=251.72022.01 | 28.3 | — | |
| SlowFast, 16x8 R101+NLPre-train=K600, GFLOPs=296.3, Param (M)=59.22022.01 | 27.5 | — | |
| MViTv1-B, 32x3Pre-train=K600, GFLOPs=169.8, Param (M)=36.42022.01 | 27.5 | — | |
| X3D-XLPre-train=K600, GFLOPs=48.4, Param (M)=112022.01 | 27.4 | — | |
| MViTv1-B, 64x3Pre-train=K400, GFLOPs=454.7, Param (M)=36.42022.01 | 27.3 | — | |
| SlowFast, 8x8 R101+NLPre-train=K600, GFLOPs=146.6, Param (M)=59.22022.01 | 27.1 | — | |
| MViTv1-B, 32x3Pre-train=K400, GFLOPs=169.8, Param (M)=36.42022.01 | 26.8 | — | |
| MViT-16, 16x4Pre-train=K400, GFLOPs=57.4, Param (M)=34.52022.01 | 26.2 | 27 | |
| MViTv1-B, 16x4Pre-train=K600, GFLOPs=70.4, Param (M)=36.32022.01 | 26.1 | — | |
| WOO, SFR50Pre-train=K400, GFLOPs=147.52022.01 | 25.4 | — | |
| MViTv1-B, 16x4Pre-train=K400, GFLOPs=70.5, Param (M)=36.42022.01 | 24.5 | — | |
| SlowFast, 8x8, R101Pre-train=K400, GFLOPs=137.7, Param (M)=532022.01 | 23.8 | — | |
| SlowFast, 8x8, R50Pre-train=K400, GFLOPs=96.9, Param (M)=33.82022.01 | 22.7 | — | |
| SlowFast, 4x16, R50Pre-train=K400, GFLOPs=52.6, Param (M)=33.72022.01 | 21.9 | — | |
| ACAR R101Pre-train=K700, GFLOPs=212, Param (M)=107.4+2022.01 | — | 33.3 |