CrossEarth-SAR: A SAR-Centric and Billion-Scale Geospatial Foundation Model for Domain Generalizable Semantic Segmentation

TL;DR

CrossEarth-SAR is a billion-scale SAR foundation model using a physics-guided MoE architecture, designed to improve semantic segmentation across diverse domains and sensors, achieving state-of-the-art results.

Contribution

The paper introduces the first billion-scale SAR foundation model with a physics-guided MoE architecture and a comprehensive benchmark suite for domain generalization.

Findings

Achieves state-of-the-art results on 20 benchmarks.

Surpasses previous methods by over 10% mIoU on some benchmarks.

Develops a large-scale SAR dataset and benchmark suite.

Abstract

Synthetic Aperture Radar (SAR) enables global, all-weather earth observation. However, owing to diverse imaging mechanisms, domain shifts across sensors and regions severely hinder its semantic generalization. To address this, we present CrossEarth-SAR, the first billion-scale SAR vision foundation model built upon a novel physics-guided sparse mixture-of-experts (MoE) architecture incorporating physical descriptors, explicitly designed for cross-domain semantic segmentation. To facilitate large-scale pre-training, we develop CrossEarth-SAR-200K, a weakly and fully supervised dataset that unifies public and private SAR imagery. We also introduce a benchmark suite comprising 22 sub-benchmarks across 8 distinct domain gaps, establishing the first unified standard for domain generalization semantic segmentation on SAR imagery. Extensive experiments demonstrate that CrossEarth-SAR achieves state-of-the-art results on 20 benchmarks, surpassing previous methods by over 10\% mIoU on some benchmarks under multi-gap transfer. All code, benchmark and datasets will be publicly available.