Action Step Localization on CrossTask (test)
52.5RecallTempCLR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TempCLREvaluation Protocol=Supervised2022.12 | 52.5 | — | |
| VideoCLIPEvaluation Protocol=Supervised2022.12 | 47.3 | — | |
| VLMAlignment Protocol=Pairwise Matching, Supervision Level=supervised on 540 videos2021.05 | 46.5 | — | |
| TempCLREvaluation Protocol=Few-shot, Training Data Scale=20%2022.12 | 42.8 | — | |
| UniVLfine-tuned=false2020.02 | 42 | — | |
| UniVLAlignment Protocol=Joint Alignment2021.05 | 42 | — | |
| UniVLEvaluation Protocol=Zero-shot2022.12 | 42 | — | |
| ActBERT2020.02 | 41.4 | — | |
| ActBERTAlignment Protocol=Pairwise Matching2021.05 | 41.4 | — | |
| VideoCLIPEvaluation Protocol=Few-shot, Training Data Scale=20%2022.12 | 41.1 | — | |
| VT-TWINSEvaluation Protocol=Zero-shot2022.12 | 40.7 | — | |
| MIL-NCE2020.02 | 40.5 | — | |
| MIL-NCEAlignment Protocol=Joint Alignment2021.05 | 40.5 | — | |
| MIL-NCEEvaluation Protocol=Zero-shot2022.12 | 40.5 | — | |
| TempCLREvaluation Protocol=Zero-shot, TFS=true2022.12 | 36.9 | — | |
| DWSAEvaluation Protocol=Zero-shot2022.12 | 35.3 | — | |
| MCNEvaluation Protocol=Zero-shot2022.12 | 35.1 | — | |
| VideoCLIPEvaluation Protocol=Zero-shot2022.12 | 33.9 | — | |
| HowTo100M2020.02 | 33.6 | — | |
| HowTo100MAlignment Protocol=Joint Alignment2021.05 | 33.6 | — | |
| HT100MEvaluation Protocol=Zero-shot2022.12 | 33.6 | — | |
| VideoCLIPEvaluation Protocol=Zero-shot, TFS=true2022.12 | 33.5 | — | |
| Supervisedmode=supervised2020.02 | 31.6 | — | |
| SupervisedAlignment Protocol=Joint Alignment2021.05 | 31.6 | — | |
| SupervisedEvaluation Protocol=Supervised2022.12 | 31.6 | — | |
| VLMAlignment Protocol=Pairwise Matching, Supervision Level=task-agnostic, zero-shot2021.05 | 28.5 | — | |
| Zhukov et al.2020.02 | 22.4 | — | |
| ZhukovAlignment Protocol=Joint Alignment2021.05 | 22.4 | — | |
| ZhukovEvaluation Protocol=Supervised2022.12 | 22.4 | — | |
| Alayrac et al.2020.02 | 13.3 | — | |
| AlayracAlignment Protocol=Joint Alignment2021.05 | 13.3 | — | |
| AlayracEvaluation Protocol=Supervised2022.12 | 13.3 | — | |
| ActBERTLabeled Dataset=CT2022.03 | — | 37.1 | |
| Alayrac et al.Labels used=ImNet+K4002019.12 | — | 13.3 | |
| Alayrac et al.Labeled Dataset=IM, K4002022.03 | — | 13.3 | |
| CrossTaskLabels used=ImNet+K4002019.12 | — | 22.4 | |
| CrossTaskLabels used=ImNet+K400+CT2019.12 | — | 31.6 | |
| CrossTaskLabeled Dataset=IM, K4002022.03 | — | 22.4 | |
| CrossTaskLabeled Dataset=IM, K400, CT2022.03 | — | 31.6 | |
| DWSALabeled Dataset=CT2022.03 | — | 35.5 | |
| Miech et al.Labels used=ImNet+K4002019.12 | — | 33.6 | |
| Miech et al.Labeled Dataset=IM, K4002022.03 | — | 33.6 | |
| MIL-NCENet=I3D, Labels used=None2019.12 | — | 36.4 | |
| MIL-NCENet=S3D, Labels used=None2019.12 | — | 40.5 | |
| MIL-NCELabeled Dataset=None2022.03 | — | 35.5 | |
| VT-TWINSLabeled Dataset=None2022.03 | — | 40.7 |