Tile2Vec: Unsupervised representation learning for spatially distributed data
About
Geospatial analysis lacks methods like the word vector representations and pre-trained networks that significantly boost performance across a wide range of natural language and computer vision tasks. To fill this gap, we introduce Tile2Vec, an unsupervised representation learning algorithm that extends the distributional hypothesis from natural language -- words appearing in similar contexts tend to have similar meanings -- to spatially distributed data. We demonstrate empirically that Tile2Vec learns semantically meaningful representations on three datasets. Our learned representations significantly improve performance in downstream classification tasks and, similar to word vectors, visual analogies can be obtained via simple arithmetic in the latent space.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Image Classification | EuroSAT | Accuracy73.39 | 569 | |
| Socio-economic indicator estimation | Cambodia socio-economic dataset | RMSE (POP)0.954 | 11 | |
| Socio-economic indicator estimation | Vietnam socio-economic dataset | RMSE (GRDP)0.875 | 11 | |
| Socio-economic indicator estimation | South Korea socio-economic dataset | RMSE (GRDP)1.342 | 11 | |
| Socioeconomic indicator estimation | Vietnam (test) | GRDP0.418 | 11 | |
| Socioeconomic indicator estimation | South Korea (test) | GRDP Estimate Score32.7 | 11 | |
| Socioeconomic indicator estimation | Cambodia (test) | POP0.621 | 11 | |
| Classification | CashewPlant | Accuracy34.4 | 8 | |
| Classification | BEN V2 | Accuracy54.95 | 8 | |
| Classification | S4A-tiles | Accuracy35.2 | 8 |