Egocentric Text Retrieval on Nymeria
32.6Top-1 Accuracy (legs)Gemini-2.5-Pro
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Gemini-2.5-ProModality=video, Zero-shot Evaluation=true2025.11 | 32.6 | 36.6 | 60.7 | 52.7 | 45.6 | |
| CamFormer embeddingsModality=trajectory, # MACs (G)=0.02, # Params (M)=0.3, Zero-shot Evaluation=true2025.11 | 30.8 | 33.2 | 36.3 | 26 | 31.6 | |
| CamFormer embeddings*Modality=video+trajectory, # MACs (G)=89.51, # Params (M)=151.0, Zero-shot Evaluation=true2025.11 | 30.1 | 30.6 | 47.8 | 34 | 35.6 | |
| EgoVLPv2 (Ego-Exo4D)Modality=video, # MACs (G)=89.49, # Params (M)=150.7, Zero-shot Evaluation=true2025.11 | 22.8 | 20.5 | 39.6 | 28.5 | 27.9 | |
| CLIPModality=image, # MACs (G)=2.95, # Params (M)=59.0, Zero-shot Evaluation=true2025.11 | 22.4 | 22.4 | 40.5 | 29.5 | 28.7 | |
| EgoVLPv2 (Ego4D)Modality=video, # MACs (G)=89.49, # Params (M)=150.7, Zero-shot Evaluation=true2025.11 | 22.1 | 17.4 | 45.7 | 31.7 | 29.2 |