Open-domain Visual Entity Recognition on OVEN Wiki (human evaluation set)
76.1Score (Seen Entities)Human+Search
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Human+SearchNote=Human performance with retrieval tools2023.02 | 76.1 | 79.3 | 77.7 | |
| PaLI-17B# Params=17B, Training Type=Fine-tuned2023.02 | 40.3 | 26 | 31.6 | |
| PaLI-3B# Params=3B, Training Type=Fine-tuned2023.02 | 30.5 | 15.8 | 20.8 | |
| CLIP Fusion ViT-L/14# Params=0.88B, Training Type=Fine-tuned2023.02 | 18 | 2.9 | 5 | |
| CLIP2CLIP ViT-L/14# Params=0.86B, Training Type=Fine-tuned2023.02 | 14 | 11.1 | 12.4 | |
| CLIP ViT-L/14# Params=0.42B, Training Type=Fine-tuned2023.02 | 4.6 | 6 | 5.2 |