Contextual Image Generation on VisDial 1.0 (test)
0.645CLIP SimilarityGILL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GILLNumber of dialogue rounds=102023.05 | 0.645 | 0.714 | |
| Stable DiffusionNumber of dialogue rounds=52023.05 | 0.629 | 0.722 | |
| Stable DiffusionNumber of dialogue rounds=102023.05 | 0.622 | 0.723 | |
| GILLNumber of dialogue rounds=52023.05 | 0.621 | 0.718 | |
| GLIDENumber of dialogue rounds=52023.05 | 0.595 | 0.794 | |
| GLIDENumber of dialogue rounds=102023.05 | 0.587 | 0.799 | |
| GLIDENumber of dialogue rounds=12023.05 | 0.562 | 0.8 | |
| Stable DiffusionNumber of dialogue rounds=12023.05 | 0.552 | 0.742 | |
| GILLNumber of dialogue rounds=12023.05 | 0.528 | 0.742 |