Grounded Open Vocabulary Acquisition on Flickr30k Entities 1.0 (Unseen words (Vunseen = 31))
32.7HR@1ViLT
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| ViLTFine-tuned=false2023.06 | 32.7 | — | — | — | — | 3.68 | — | — | — | — | |
| ViLT+MDETRFine-tuned=false2023.06 | 32.7 | 8.6 | 8.1 | 5.07 | 5.12 | 3.68 | 27.3 | 23.3 | 25 | 23.8 | |
| ROBERTaFine-tuned=true2023.06 | 24.3 | — | — | — | — | 4.38 | — | — | — | — | |
| ROBERTaFine-tuned=false2023.06 | 23.1 | — | — | — | — | 4.96 | — | — | — | — | |
| VisualBERTFine-tuned=true2023.06 | 20.7 | 10.2 | — | 5.6 | — | 4.81 | 50.6 | — | 45.2 | — | |
| OctoBERTFine-tuned=false2023.06 | 4.2 | 2.3 | 2.3 | 11.58 | 11.74 | 11.01 | 61.3 | 53.1 | 56.3 | 48 | |
| OctoBERT_w/o GFine-tuned=true2023.06 | 3.7 | 1.1 | 1.1 | 11.89 | 12.04 | 10.87 | 38.7 | 31.9 | 36.2 | 31 | |
| MDETRFine-tuned=false2023.06 | — | — | — | — | — | — | 26.3 | 20.2 | 23.9 | 21.7 |