Image-Text Consistency on Urban-1k (test)
47.25Consistency (Word Dropout p=0.5)GT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GTDescription=Ground-truth image paired with original caption, Evaluation Framework=BLIP-VQA, Runs=52026.07 | 47.25 | 48.69 | 48.09 | 45.81 | 43.95 | |
| HyFL-CLIPSelection=Top-1 retrieved image, Evaluation Framework=BLIP-VQA, Runs=52026.07 | 41.12 | 42.83 | 41.6 | 43.72 | 39.97 | |
| HiMo-CLIPSelection=Top-1 retrieved image, Evaluation Framework=BLIP-VQA, Runs=52026.07 | 38.64 | 39.82 | 39.17 | 43.21 | 37.95 |