TTE / README.md
dcher95's picture
Mirror weights from dcher95/TTE; update model card
488591e verified
|
Raw
History Blame Contribute Delete
1.22 kB
metadata
library_name: tte
pipeline_tag: feature-extraction
tags:
  - location-encoder
  - geospatial
  - remote-sensing
  - sentinel-2
  - voronoi
license: mit

Tessellating the Earth (TTE) — location encoder

Maps a geographic coordinate (lat, lon) to a learned embedding via a learnable Spherical Voronoi partition of S² with global semantic tokens. ECCV 2026.

Daniel Cher, Hamza Iqbal, Eric Xing, Brian Wei, Nathan Jacobs — Washington University in St. Louis (MVRL).

import torch
from tte import TTE

model = TTE.from_pretrained("MVRL/TTE").eval()
coords = torch.tensor([[37.77, -122.42],
                       [-3.12,   60.02]])
emb = model.encode(coords)

Image backbone used during training (not needed here): frozen SSL4EO-S12 MAE ViT-L/16.

Citation

@inproceedings{cher2026tte,
  title     = {Tessellating the Earth: Learnable Spherical Voronoi Partitions for Location Encoding},
  author    = {Cher, Daniel and Iqbal, Hamza and Xing, Eric and Wei, Brian and Jacobs, Nathan},
  booktitle = {European Conference on Computer Vision (ECCV)},
  year      = {2026}
}