Grounded Open Vocabulary Acquisition on Flickr30k Entities 1.0 (Seen Vseen = 60)
66.9HR@1OctoBERT
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| OctoBERTFine-tuned=false2023.06 | 66.9 | 47 | 46.3 | 1.79 | 1.81 | 1.26 | 66.8 | 66.3 | 58.8 | 57.6 | |
| ViLTFine-tuned=false2023.06 | 64.7 | — | — | — | — | 1.27 | — | — | — | — | |
| ViLT+MDETRFine-tuned=false2023.06 | 64.7 | 19.8 | 19.3 | 2.53 | 2.43 | 1.27 | 31.1 | 30.4 | 28.5 | 31.2 | |
| OctoBERT_w/o GFine-tuned=true2023.06 | 63.9 | 28.9 | 27.8 | 2.33 | 2.38 | 1.41 | 44 | 43 | 40 | 38.2 | |
| ROBERTaFine-tuned=true2023.06 | 47.9 | — | — | — | — | 1.99 | — | — | — | — | |
| VisualBERTFine-tuned=true2023.06 | 42.3 | 28.5 | — | 2.96 | — | 2.33 | 68.1 | — | 53.3 | — | |
| ROBERTaFine-tuned=false2023.06 | 38 | — | — | — | — | 2.75 | — | — | — | — | |
| MDETRFine-tuned=false2023.06 | — | — | — | — | — | — | 27.8 | 27 | 25.3 | 28 |