Expert Demonstration Retrieval on Ego-Exo4D 1.0 (test)
22.5Recall@50ExpertAF w/ full-sup
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ExpertAF w/ full-supprivileged_input=demo video at inference2024.08 | 22.5 | 146 | — | |
| ExpertAF2024.08 | 19.1 | 158 | 3.2 | |
| ExpertAF w/o videoablation=no video2024.08 | 18.7 | 161 | — | |
| ExpertAF w/o poseablation=no pose2024.08 | 18.6 | 163 | — | |
| LLaVA-FT w/ posemode=Fine-tuned, pose_input=true2024.08 | 18 | 172 | 2.6 | |
| LLaVA-FTmode=Fine-tuned2024.08 | 17.8 | 177 | — | |
| ExpertAF w/o alignmentablation=no temporal alignment2024.08 | 16.9 | 180 | — | |
| ExpertAF w/ global poseablation=global pose2024.08 | 16.5 | 184 | — | |
| PoseScript/Fix2024.08 | 15.9 | 182 | 1.6 | |
| LLaVA2024.08 | 15 | 183 | 1.2 | |
| VideoChat22024.08 | 14.9 | 183 | — | |
| InternVideo2-FTmode=Fine-tuned2024.08 | 14.1 | 190 | 2.1 | |
| InternVideo2-NNmode=Nearest Neighbor2024.08 | 13.5 | 198 | — |