Question Generation on MSCOCO-VQA (test)
60.8METEORHumans2016
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Humans20162018.08 | 60.8 | — | — | 86 | — | — | — | — | |
| MDN-JointModule=Joint Mixture2018.08 | 23.4 | 50.7 | — | 36 | — | — | — | 41.8 | |
| Creative20172018.08 | 19.9 | — | — | 35.6 | — | — | — | — | |
| Natural20162018.08 | 19.7 | — | — | 19.2 | — | — | — | — | |
| Diff.Image-JointType=Baseline ablation2018.08 | 11.7 | 38.8 | — | 30.4 | — | — | — | 26.3 | |
| Tag-HadamardType=Baseline ablation2018.08 | 10.8 | 55 | — | 24.4 | — | — | — | 24.3 | |
| PlaceCNN-JointType=Baseline ablation2018.08 | 10.8 | 56.1 | — | 25.7 | — | — | — | 24.5 | |
| Image OnlyInput=Image2018.08 | 8.6 | 18.8 | — | 20.8 | — | — | — | 22.6 | |
| Caption OnlyInput=Caption2018.08 | 8.5 | 22.3 | — | 21.1 | — | — | — | 25.9 | |
| coco-Caption2015.12 | 0.195 | 0.66 | — | 0.625 | 0.45 | 0.321 | 0.23 | — | |
| Neural Self TalkGeneration strategy=MAX (max probability from trained model)2015.12 | 0.178 | 0.331 | 0.493 | 0.594 | 0.422 | 0.291 | 0.193 | — | |
| Neural Self TalkGeneration strategy=SAMPLE (randomly drawn from trained probabilistic model)2015.12 | 0.127 | 0.133 | 0.342 | 0.388 | 0.22 | 0.117 | 0.064 | — |