Image-to-Text Retrieval on Flickr (non-Karpathy)
80Top-1 Recall3T
Evaluation results
| Method | Links | |
|---|---|---|
| 3TPretraining=IN-21k, Model Scale=L, Training Dataset=Unfiltered WebLI, Evaluation Protocol=Zero-shot2023.05 | 80 | |
| 3TPretraining=JFT, Model Scale=L, Training Dataset=Unfiltered WebLI, Evaluation Protocol=Zero-shot2023.05 | 80 | |
| LiTPretraining=JFT, Model Scale=L, Training Dataset=Unfiltered WebLI, Evaluation Protocol=Zero-shot2023.05 | 78.7 | |
| Baseline (CLIP/ALIGN)Pretraining=IN-21k, Model Scale=L, Training Dataset=Unfiltered WebLI, Evaluation Protocol=Zero-shot2023.05 | 75.6 | |
| LiTPretraining=IN-21k, Model Scale=L, Training Dataset=Unfiltered WebLI, Evaluation Protocol=Zero-shot2023.05 | 71.7 |