Text-to-Image Retrieval on Flickr (non-Karpathy)
61.4Top-1 Recall3T
Evaluation results
| Method | Links | |
|---|---|---|
| 3TPretraining=JFT, Model Scale=L, Training Dataset=Unfiltered WebLI, Evaluation Protocol=Zero-shot2023.05 | 61.4 | |
| 3TPretraining=IN-21k, Model Scale=L, Training Dataset=Unfiltered WebLI, Evaluation Protocol=Zero-shot2023.05 | 60.9 | |
| LiTPretraining=JFT, Model Scale=L, Training Dataset=Unfiltered WebLI, Evaluation Protocol=Zero-shot2023.05 | 58.8 | |
| Baseline (CLIP/ALIGN)Pretraining=IN-21k, Model Scale=L, Training Dataset=Unfiltered WebLI, Evaluation Protocol=Zero-shot2023.05 | 57.1 | |
| LiTPretraining=IN-21k, Model Scale=L, Training Dataset=Unfiltered WebLI, Evaluation Protocol=Zero-shot2023.05 | 49.3 |