Multimodal Machine Translation (English-German) on MSCOCO (test)
31.14BLEUSelective Attn + ViT-Base
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Selective Attn + ViT-BaseFeature=ViT-Base2022.03 | 31.14 | 56.77 | |
| Selective AttnFeature=ViT-Large2022.03 | 30.22 | 56.91 | |
| Selective Attn + QueryInstFeature=QueryInst2022.03 | 30.2 | 56.89 | |
| Selective Attn + DETRFeature=DETR2022.03 | 30.13 | 57.01 | |
| ImaginationFeature=ResNet2022.03 | 29.9 | 56.57 | |
| Transformer TinyFeature=Text-only2022.03 | 29.88 | 56.64 | |
| Selective Attn + ViT-SmallFeature=ViT-Small2022.03 | 29.72 | 56.94 | |
| Doubly-ATTFeature=ResNet2022.03 | 29.63 | 56.21 | |
| Selective Attn + CATRFeature=CATR2022.03 | 29.59 | 56.36 | |
| Gated FusionFeature=ViT-Large2022.03 | 29.27 | 55.64 | |
| Gated FusionFeature=ResNet2022.03 | 29.04 | 56.15 | |
| UVR-NMTFeature=ResNet2022.03 | 29.02 | — | |
| Selective Attn + ViT-TinyFeature=ViT-Tiny2022.03 | 28.1 | 55.19 |