M4-SAR
收藏资源简介:
M4-SAR是一个多分辨率、多极化、多场景、多源数据集,用于光学与合成孔径雷达(SAR)融合的目标检测。该数据集由南京理工大学PCA实验室、安徽大学ICSP教育部重点实验室和南开大学计算机科学学院共同构建,包含112,184对精确对齐的图像和近一百万个标注实例。数据集覆盖六个关键类别,并使用公开的光学和SAR数据,包括Sentinel-1和Sentinel-2卫星提供的数据。为了克服SAR标注的挑战,该研究提出了一种半监督的光学辅助标注策略,利用光学图像的语义丰富性来显著提高标注质量。M4-SAR数据集旨在解决现有光学和SAR数据集的局限性,为多源融合目标检测任务提供大规模、高质量、标准化的数据集,并推动相关研究的发展。
M4-SAR is a multi-resolution, multi-polarization, multi-scene, and multi-source dataset for object detection via optical and Synthetic Aperture Radar (SAR) fusion. This dataset was jointly constructed by the PCA Laboratory of Nanjing University of Science and Technology, the Key Laboratory of ICSP (Ministry of Education) of Anhui University, and the School of Computer Science of Nankai University. It contains 112,184 precisely aligned image pairs and nearly one million annotated instances. The dataset covers six key categories and utilizes publicly available optical and SAR data, including data provided by Sentinel-1 and Sentinel-2 satellites. To address the challenges of SAR annotation, this study proposes a semi-supervised optical-assisted annotation strategy that leverages the semantic richness of optical images to significantly improve annotation quality. The M4-SAR dataset aims to resolve the limitations of existing optical and SAR datasets, provide a large-scale, high-quality, standardized dataset for multi-source fusion object detection tasks, and advance the development of related research.




