Semantic Role Prediction on VidSitu (test)
84.85CIDErHuman Level
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Human LevelGT verb and role pairs=true2022.10 | 84.85 | 91.7 | 80.15 | 39.77 | 70.33 | — | — | |
| Human LevelGT verb and role pairs=true2022.10 | 83.68 | 87.78 | 79.29 | 40.04 | 71.77 | — | — | |
| VideoWhispererGT verb and role pairs=true2022.10 | 68.54 | 77.48 | 61.55 | 45.7 | 47.54 | — | — | |
| VideoWhispererGT verb and role pairs=true2022.10 | 68.04 | 81.23 | 62.19 | 46.15 | 48.77 | — | — | |
| Slow-D+TxE+TxDbackbone pretraining=CVRL-K400, TxE pretraining=MaskPred-LVU2022.04 | 61.18 | 69.15 | 54.99 | 43.38 | 47.81 | — | — | |
| Slow-D+TxE+TxDbackbone pretraining=CVRL-K400, TxE pretraining=MaskPred-VS2022.04 | 60.34 | 69.12 | 53.87 | 43.77 | 46.77 | — | — | |
| Slow-D+TxE+TxDGT verb and role pairs=true2022.10 | 60.34 | 69.12 | 53.87 | 43.77 | 46.77 | — | — | |
| Slow-D+TxE+TxDbackbone pretraining=CVRL-K400, TxE pretraining=MaskPred-K4002022.04 | 57.48 | 65.08 | 51.21 | 41.65 | 45.71 | — | — | |
| Slow-D+TxE+TxDbackbone pretraining=CVRL-K400, TxE pretraining=N (not pretrained)2022.04 | 54.4 | 63.18 | 47.63 | 41.8 | 46.31 | — | — | |
| Slow-D+TxDbackbone pretraining=Sup-K400, TxE pretraining=none2022.04 | 51.37 | 59.68 | 46.1 | 41.37 | 36.03 | — | — | |
| Slow-D+TxE+TxDbackbone pretraining=Sup-K400, TxE pretraining=N (not pretrained)2022.04 | 51.36 | 59.72 | 47.25 | 41.72 | 45.99 | — | — | |
| SlowFast+TxE+TxDGT verb and role pairs=true2022.10 | 47.25 | 52.92 | 45.48 | 43.46 | 50.88 | — | — | |
| I3D+TxDbackbone pretraining=Sup-K400, TxE pretraining=none2022.04 | 47.14 | 51.61 | 41.29 | 40.67 | 37.89 | — | — | |
| I3D+TxE+TxDbackbone pretraining=Sup-K400, TxE pretraining=N (not pretrained)2022.04 | 47.06 | 51.67 | 42.76 | 42.41 | 48.92 | — | — | |
| SlowFast+TxE+TxDGT verb and role pairs=true2022.10 | 46.01 | 56.37 | 43.58 | 43.04 | 50.89 | — | — | |
| Slow-D+TxE+TxDbackbone pretraining=CVRL-VS, TxE pretraining=MaskPred-VS2022.04 | 44.32 | 52.07 | 39.08 | 40.56 | 48.87 | — | — | |
| GPT2backbone pretraining=visual-blind, TxE pretraining=none2022.04 | 34.67 | 42.97 | 34.45 | 40.08 | 48.08 | — | — |