Visual Entity Recognition on OVEN (test)
33.6Top-1 Acc (Seen)CLIP-L/14 Fusion
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CLIP-L/14 Fusion# params (M)=880, Evaluation Protocol=Fine-tuning on the OVEN Seen categories2023.06 | 33.6 | 4.8 | 8.4 | |
| PaLI-17B# params (M)=17000, Evaluation Protocol=Fine-tuning on the OVEN Seen categories2023.06 | 28.3 | 11.2 | 16.1 | |
| PaLI-3B# params (M)=3000, Evaluation Protocol=Fine-tuning on the OVEN Seen categories2023.06 | 19.1 | 6 | 9.3 | |
| CLIP-L/14 CLIP2CLIP# params (M)=860, Evaluation Protocol=Fine-tuning on the OVEN Seen categories2023.06 | 12.6 | 10.5 | 11.5 | |
| CLIP-L/14 + RECO# params (M)=435, Evaluation Protocol=Zero-shot2023.06 | 11.5 | 13.3 | 12.3 | |
| CLIP-L/14# params (M)=428, Evaluation Protocol=Zero-shot2023.06 | 5.6 | 4.9 | 5.3 | |
| PaLI-17B# params (M)=17000, Evaluation Protocol=Zero-shot2023.06 | 4.4 | 1.2 | 1.9 |