Touchdown: Natural Language Navigation and Spatial Reasoning in Visual Street Environments
Howard Chen, Alane Suhr, Dipendra Misra, Noah Snavely, Yoav Artzi
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/lil-lab/touchdownOfficialIn paperpytorch★ 109
- github.com/VegB/VLN-Transformerpytorch★ 27
- github.com/clic-lab/ciffpytorch★ 0
- github.com/lil-lab/ciffpytorch★ 0
Abstract
We study the problem of jointly reasoning about language and vision through a navigation and spatial reasoning task. We introduce the Touchdown task and dataset, where an agent must first follow navigation instructions in a real-life visual urban environment, and then identify a location described in natural language to find a hidden object at the goal position. The data contains 9,326 examples of English instructions and spatial descriptions paired with demonstrations. Empirical analysis shows the data presents an open challenge to existing methods, and qualitative linguistic analysis shows that the data displays richer use of spatial reasoning compared to related resources.
Tasks
Benchmark Results
| Dataset | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| Touchdown Dataset | Gated Attention (GA) | Task Completion (TC) | 5.5 | — | Unverified |
| Touchdown Dataset | RConcat | Task Completion (TC) | 10.7 | — | Unverified |
| Touchdown Dataset | Gated Attention (GA) | Task Completion (TC) | 11.9 | — | Unverified |
| Touchdown Dataset | RConcat | Task Completion (TC) | 11.8 | — | Unverified |