AirMOS: A Global Optical-SAR Multimodal Benchmark for Airport Detection

AirMOS is a global benchmark dataset for multimodal remote sensing airport object detection, designed to provide a unified and standardized evaluation platform for optical–SAR fusion detection research. Covering representative airport scenes around the world, the dataset features strong geographic diversity and scene variability, making it well suited for evaluating model detection performance and generalization ability in complex environments.AirMOS contains 4,517 high-quality multimodal sample pairs, integrating Sentinel-2 optical imagery with Sentinel-1 dual-polarization SAR data to form a unified three-modality data framework consisting of optical, SAR VV, and SAR VH. The dataset supports both unimodal and multimodal object detection tasks, providing a reliable benchmark for comparing the performance of different detection models and fusion strategies.To further assess algorithm robustness under degraded observation conditions, AirMOS also provides cloud- and haze-degraded versions generated using a fractal-noise-based augmentation method. This approach enables controllable simulation of cloud morphology, spatial distribution density, and occlusion intensity, and produces two levels of visible-image degradation: moderate occlusion (L1) and extreme occlusion (L2). These degraded subsets offer valuable support for multimodal detection research under adverse weather conditions.