GEOID-Flood benchmarks flood segmentation models
The dataset spans 219 flood events across 65 countries using Sentinel imagery, DEM data and validated labels.
Why it matters
Flood mapping is a practical test of whether geospatial foundation models transfer across regions and sensors. GEOID-Flood gives researchers a larger benchmark for comparing optical-SAR fusion, fine-tuning and conventional approaches on a disaster-response task.
The key points
- 1.GEOID-Flood spans 219 events in 65 countries.
- 2.It combines Sentinel-1, Sentinel-2, DEM data and validated labels.
- 3.Foundation models showed consistent but modest gains.
Researchers introduced GEOID-Flood, a large-scale multimodal benchmark for flood segmentation derived from Copernicus Emergency Management Service activations. The dataset covers 219 events across 65 countries over ten years and includes more than 14,000 tiles with co-registered pre- and post-event Sentinel-1, pre-event Sentinel-2 composites, DEM data and manually validated labels distinguishing background, permanent water and flooded water. The benchmark evaluates foundation models and conventional encoders across single-image, multi-temporal and multi-modal protocols.
⚡ Try this today
Use GEOID-Flood to benchmark flood segmentation models before deploying geospatial foundation models for flood mapping.
Sources & original reporting
This brief summarizes and links to reporting from the publishers below.
Enjoyed this brief? Get the next one in your inbox.
More in Research
Prior Audit-Repair Context Shifts LLM Verifier Thresholds Toward Leniency
arXiv:2608.
Improving the matrix multiplication exponent with modern optimization and AlphaEvolve
arXiv:2608.
[Paper] Intern-S2-Mobius: Foundation Model with Decoupled Knowledge and Reasoning
We introduce Mobius-v0, an architecture that comprises a globally shared Memory (FFN) that stores knowledge vectors and multiple Reasoners (Self-Attn) that iter