Text-to-image synthesis evaluation (Human Correlation) on COCO 2017
0.3357Kendall's TauLLMScore
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LLMScoreHuman ranking criteria=Overall2023.05 | 0.3357 | 0.4275 | |
| NegCLIPHuman ranking criteria=Overall2023.05 | 0.2793 | 0.3523 | |
| CLIPHuman ranking criteria=Overall2023.05 | 0.2227 | 0.2771 | |
| LLMScoreHuman ranking criteria=Error Counting2023.05 | 0.2038 | 0.3027 | |
| CLIPHuman ranking criteria=Error Counting2023.05 | 0.1888 | 0.2677 | |
| BLIP-ITCHuman ranking criteria=Error Counting2023.05 | 0.1849 | 0.262 | |
| NegCLIPHuman ranking criteria=Error Counting2023.05 | 0.1795 | 0.2581 | |
| BLIP-ITCHuman ranking criteria=Overall2023.05 | 0.1703 | 0.2121 | |
| BLIP-ITMHuman ranking criteria=Error Counting2023.05 | 0.1137 | 0.1635 | |
| BLIP-ITMHuman ranking criteria=Overall2023.05 | 0.0928 | 0.1155 |