A dataset of multimodal visible-SAR imagery for airport detection in remote sensing

<p indent="0mm">Remote sensing image object detection is a significant area in Earth observation tasks. By integrating remote sensing data from various sensors and utilizing multimodal complementary information, the performance of object detection models can be significantly improved. To address the current limitations in the remote sensing multimodal object detection field, such as insufficient datasets, limited modality combinations, and restricted object categories, this study constructs a multimodal Visible-SAR dataset for airport detection in remote sensing imagery named MVSDA, which is based on Google Earth visible images and Gaofen-3 (GF-3) Synthetic Aperture Radar (SAR) imagy. During dataset construction, standardized protocols were followed for data collection, processing, and annotation, encompassing a series of procedures such as multi-source data acquisition, spatial alignment, non-informative region filtering, expert-guided annotation, quality control, and evaluation. This dataset is characterized by pixel-level alignment, large geographic coverage, multiple spatial resolutions, and diverse environmental conditions, which supports the training requirements of mainstream deep learning models. This dataset provides essential foundational resources for researchers in related fields and holds practical value for advancing research in remote sensing multimodal object detection.