google/RSRCC

RSRCC（通过检索增强的最佳排名构建的遥感区域变化理解基准）是一个用于遥感语义变化理解的数据集，包含多时相图像对和自然语言问题及回答。数据集旨在支持对遥感场景中时间变化的更丰富理解，包括新建、拆除、道路或人行道变化、植被变化和住宅开发等示例。数据集具有语义变化理解、图像对推理、指令式注释、多样化问题格式和遥感焦点等关键特点。数据集结构分为训练、验证和测试三个标准分割，每个分割包含图像文件夹和元数据文件。每个样本包括变化前后的图像以及描述语义变化的自然语言问题和回答。数据集适用于语义变化描述、遥感图像上的视觉语言推理、多模态问答、时间场景理解和遥感基础模型的指令调优等研究。

RSRCC (A Remote Sensing Regional Change Comprehension Benchmark Constructed via Retrieval-Augmented Best-of-N Ranking) is a dataset designed for semantic change understanding in remote sensing, pairing multi-temporal image evidence with natural language questions and answers. The dataset was created to support this richer understanding of temporal change in remote sensing scenes, including examples such as new construction, demolition, road or sidewalk changes, vegetation changes, and residential development. Key features include semantic change understanding, image-pair reasoning, instruction-style annotations, diverse question formats, and a remote sensing focus. The dataset is organized into three standard splits (train, val, test), each containing image folders and a metadata file. Each sample includes a before image, an after image, and a natural-language question and answer describing the semantic change. The dataset is intended for research on semantic change captioning, vision-language reasoning over remote sensing imagery, multimodal question answering, temporal scene understanding, and instruction tuning for remote sensing foundation models.