Controlled Caption Generation on Localized Narratives of Flickr30k
64.4BLEU-1MITR
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MITRTrained on=Task 1 + Task 2 + cycle_b2021.05 | 64.4 | 37.4 | 30 | 54.7 | 2.014 | 36.5 | |
| MITR (2 layer)Trained on=Task 1 + Task 2 + cycle_b2021.05 | 64 | 36.6 | 29.7 | 54.3 | 2 | 37.4 | |
| BaselineTrained on=Task 22021.05 | 62 | 34.5 | 28.6 | 52.4 | 1.763 | 34.1 |