Temporal Action Localization on CrossTask (test)
0.414RecallActBERT
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ActBERTModality=VT, Model=+ Faster R-CNN, Trainable backbone=N2021.04 | 0.414 | — | — | — | |
| MIL-NCEModality=VT, Model=S3D-G, Trainable backbone=Y2021.04 | 0.405 | — | — | — | |
| ActBERTModality=VT, Model=R101+Res3D, Trainable backbone=N2021.04 | 0.371 | — | — | — | |
| MIL-NCEModality=VT, Model=I3D-G, Trainable backbone=Y2021.04 | 0.364 | — | — | — | |
| MCN (ours)Modality=VAT, Model=R152+RX101, Trainable backbone=N2021.04 | 0.351 | 0.336 | — | 22.2 | |
| MiechModality=VT, Model=R152+RX101, Trainable backbone=N2021.04 | 0.336 | 0.266 | — | 17.5 | |
| MIL-NCE*Modality=VT, Model=R152+RX101, Trainable backbone=N2021.04 | 0.332 | 0.302 | — | 16.3 | |
| CrossTaskModality=VT, Model=R152+I3D, Trainable backbone=N2021.04 | 0.316 | — | — | — | |
| CrossTaskModality=VT, Model=R152+I3D, Trainable backbone=N2021.04 | 0.224 | — | — | — |