BIS-1.0（SAR数据集）

SAR建筑物实例语义分割数据集-1.0 (SARMV3D-BIS-1.0) 包括复杂（D）、中等（M）和简单（S）三组数据，分别包括1024×1024大小的SAR图像1588张、773张和234张，每张都有1B级幅度图、对应的单视复图像（SLC）、叠掩次数图和标注Mask图像，涵盖标注的建筑实例分别为4.8万、1.5万和6000余个，每个建筑实例均标注了建筑立面、建筑屋顶和建筑阴影三个类别，每组数据均按照6:2:2的比例随机分成了训练集、验证集和测试集，各有一个JSON格式的标注文件记录详细的标注信息，详见该子集的说明文档。

SAR Building Instance Semantic Segmentation Dataset-1.0 (SARMV3D-BIS-1.0) includes three data groups: complex (D), moderate (M) and simple (S). There are 1588, 773 and 234 SAR images with a size of 1024×1024 for the complex, moderate and simple groups respectively. Each image is accompanied by a 1B-level amplitude image, the corresponding single-look complex (SLC) image, layover count map and annotated Mask image. The three groups respectively contain over 48,000, 15,000 and 6,000 annotated building instances, and each building instance is annotated with three categories: building facade, building roof and building shadow. Each group is randomly split into training, validation and test sets at a ratio of 6:2:2, and each of these subsets has a JSON-format annotation file that records detailed annotation information. Please refer to the documentation of the corresponding subset for more details.

搜集汇总

背景与挑战

背景概述

BIS-1.0是一个SAR建筑物实例语义分割数据集，包含复杂、中等和简单三组数据，每组提供SAR图像及对应的标注Mask图像，涵盖建筑立面、屋顶和阴影等类别。数据集已按比例划分为训练集、验证集和测试集，并附带JSON格式的标注文件。