Context assisted image captioning on NYTimes800K (test)
9.57BLEU-4NewsMEP
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| NewsMEP2023.06 | 9.57 | 13.02 | 23.62 | 65.85 | 26.61 | 28.57 | |
| Unified Vision-Language modelArchitecture=OFA-Large, Backbone=ResNet152, Visual Tokenizer=VQGAN2023.06 | 7.54 | 11.27 | 23.28 | 66.41 | 28.11 | 23.25 | |
| OFA2023.06 | 6.91 | 10.77 | 22.7 | 61.81 | 27.14 | 22.51 | |
| JOGANIC2023.06 | 6.79 | 10.93 | 22.8 | 59.42 | 28.63 | 24.49 | |
| Visual News2023.06 | 6.4 | 8.1 | 21.9 | 56.1 | 24.8 | 22.3 | |
| Transform and Tell2023.06 | 6.3 | 10.3 | 21.7 | 54.4 | 24.6 | 22.2 |