Visual Navigation on AI2-THOR Unseen Scenes (All) (test)
16.15SPLSAVN
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SAVNLoss type=learned self-supervised loss2018.12 | 16.15 | 40.86 | |
| Scene Priors2018.12 | 15.47 | 35.13 | |
| SAVN (diversity)Loss type=hand-crafted diversity loss2018.12 | 15.12 | 39.52 | |
| No adaptation (A3C)interaction-gradient=false2018.12 | 14.68 | 33.04 | |
| SAVN (prediction)Loss type=hand-crafted prediction loss2018.12 | 14.36 | 38.06 | |
| Nearest neighbor (NN)Backbone=ResNet features2018.12 | 6.09 | 7.9 | |
| Random agent2018.12 | 3.64 | 8 |