LoCA co-aligns Sentinel encoders for geospatial location encoding
Original title: LoCA: Enabling Multi-Modality in Geospatial Location Encoders
On 17 tasks LoCA wins 68.8% of per-task comparisons and is the only encoder ahead of every rival on a majority of tasks.
The authors argue Lipschitz location encoders have intrinsic dimension at most 2, so contrastive loss cannot reach its asymptotic optimum, and joint training collapses embedding covariance. LoCA first contrastively trains Sentinel-1 and Sentinel-2 encoders into a co-aligned high-rank target, then freezes them and trains location and tabular metadata encoders as multi-view adapters. Against five state-of-the-art encoders it records the most task wins.
Unconfirmed: The 68.8% win rate and task rankings are the authors' reported comparisons.
- Stage 1 jointly trains Sentinel-1 and Sentinel-2 image encoders.
- Stage 2 freezes those encoders and trains location and metadata adapters.
- Tasks include land cover, wildfire, geophysical, socio-economic, biodiversity and climate.
- LoCA is the only model ahead of every other encoder on a majority of tasks.
- Products
- Sentinel-1Sentinel-2LoCA
From the original article
1 min read at huggingface.co
Location encoders map geographic coordinates to high-dimensional descriptors through contrastive alignment with paired Earth Observation imagery. We prove that this construction is structurally limited: any Lipschitz location encoder produces an output of intrinsic dimension at most 2, regardless of architecture or bandwidth, which forces the contrastive loss to strictly exceed its asymptotic optimum.
By Panagiotis Koromilas, Eleanna Vali, Diogenis Antonopoulos · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.