Vision-and-Language Navigation on Touchdown Unseen (test)
27nDTWbest merged
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| best merged2022.03 | 27 | 19.3 | |
| ORAR mixed modelImage features=4th-to-last + no image2022.03 | 26.3 | 18.5 | |
| best non-merged2022.03 | 21.6 | 14.9 | |
| ORARImage features=ResNet 4th-to-last2022.03 | 21.6 | 14.9 | |
| ORARImage features=ResNet 4th-to-last, Heading delta=false2022.03 | 21.2 | 14.8 | |
| ORARImage features=ResNet pre-final2022.03 | 12.1 | 8.8 | |
| ORARImage features=ResNet 4th-to-last, Junction type=false2022.03 | 7.1 | 4.3 | |
| ORARImage features=ResNet 4th-to-last, Heading delta=false, Junction type=false2022.03 | 7 | 4.4 | |
| VLN Transformer2022.03 | 5.2 | 3.1 | |
| GA2022.03 | 4 | 2.2 | |
| RConcat2022.03 | 3.5 | 1.9 |