Multimodal Machine Translation on MSCOCO EN-FR (test)
45.82BLEUSelective Attn + ViT-Tiny
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Selective Attn + ViT-TinyFeature=ViT-Tiny2022.03 | 45.82 | 70.75 | |
| Gated FusionFeature=ViT-Large2022.03 | 45.65 | 70.81 | |
| Selective Attn + QueryInstFeature=QueryInst2022.03 | 45.56 | 70.64 | |
| DCCN#Params=+0.9M2020.09 | 45.4 | 65 | |
| Selective Attn + ViT-SmallFeature=ViT-Small2022.03 | 45.28 | 70.89 | |
| Selective Attn + CATRFeature=CATR2022.03 | 45.27 | 70.73 | |
| Doubly-ATTFeature=ResNet2022.03 | 45.16 | 70.25 | |
| Selective Attn + DETRFeature=DETR2022.03 | 45.1 | 70.38 | |
| Doubly-attention#Params=+3.9M2020.09 | 44.9 | 64.4 | |
| Gated FusionFeature=ResNet2022.03 | 44.86 | 70.51 | |
| Selective AttnFeature=ViT-Large2022.03 | 44.82 | 70.63 | |
| ImaginationFeature=ResNet2022.03 | 44.81 | 70.35 | |
| Selective Attn + ViT-BaseFeature=ViT-Base2022.03 | 44.72 | 71.2 | |
| Encoder-attention#Params=+1.0M2020.09 | 44.6 | 64.5 | |
| Transformer TinyFeature=Text-only2022.03 | 44.52 | 69.43 | |
| Transformer#Params=16.0M2020.09 | 44.2 | 64.2 | |
| UVR-NMTFeature=ResNet2022.03 | 43.71 | — | |
| Trg-mul#Params=-2020.09 | 43.5 | 63.2 | |
| Fusion-conv#Params=-2020.09 | 43.2 | 63.1 |