Event relation prediction on VidSitu
35.32Mean AccuracySlow-D+TxE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Slow-D+TxEbackbone pretrain=MoDist, vb finetune=true, TxE pretrain=MaskPred2022.04 | 35.32 | 41.62 | |
| Slow-D+TxEbackbone pretrain=CVRL, vb finetune=true, TxE pretrain=MaskPred2022.04 | 34.71 | 41.16 | |
| Slow-Dbackbone pretrain=MoDist, vb finetune=true2022.04 | 34.66 | 41.75 | |
| I3Dbackbone pretrain=Supervised2022.04 | 34.13 | 39.91 | |
| Slow-Dbackbone pretrain=Supervised, vb finetune=true2022.04 | 34 | 40.65 | |
| Slow-Dbackbone pretrain=CVRL, vb finetune=true2022.04 | 33.89 | 41.35 | |
| Slow-Dbackbone pretrain=MoDist, vb finetune=false2022.04 | 33.29 | 40.52 | |
| Slow-Dbackbone pretrain=Supervised, vb finetune=false2022.04 | 33.03 | 41.9 | |
| Slow-Dbackbone pretrain=CVRL, vb finetune=false2022.04 | 32.05 | 38.68 | |
| Slow-Dbackbone pretrain=No, vb finetune=true2022.04 | 30.66 | 41.29 | |
| Slow-D+TxEbackbone pretrain=CVRL, vb finetune=true, TxE pretrain=No2022.04 | 25 | 39.42 | |
| Slow-D+TxEbackbone pretrain=MoDist, vb finetune=true, TxE pretrain=No2022.04 | 25 | 39.42 |