Short-term object interaction anticipation on EGO4D v2 (test)
34.89Noun Top-5 mAPSOIA-DOD
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SOIA-DODbackbone=YOLOv9-E, visual_encoder=CLIP ViT-L/14@336px, top_k_interactions=6, inference_object_candidates=202024.07 | 34.89 | 17.61 | 10.91 | 6.22 | |
| Zarriochallenge_participant=true2024.07 | 33.5 | 17.26 | 11.77 | 6.75 | |
| EgoVideochallenge_participant=true2024.07 | 31.08 | 16.18 | 12.41 | 7.21 | |
| FRCNN+SF.2023.05 | 26.15 | 9.45 | 8.69 | 3.61 | |
| GANO v22D Backbone=ResNet-50, 3D Backbone=X3D-M2023.05 | 25.67 | 13.6 | 9.02 | 5.16 | |
| GANOv22024.07 | 25.67 | 13.6 | 9.02 | 5.16 | |
| StillFast2023.05 | 25.06 | 13.29 | 9.14 | 5.12 | |
| StillFast2024.07 | 25.06 | 13.29 | 9.14 | 5.12 |