BenthiCat - SSS-CAM

This dataset comprises co-registered optical and side-scan sonar (SSS) samples from the BenthiCat collection, designed to support multimodal learning. Each optical image is spatially associated with its corresponding SSS tile (256×256 pixels) based on geographic proximity as described in the paper. The dataset does not contain any image-level or pixel-level annotations. The dataset is organized into two primary directories: sonar and camera, which contain the matched SSS tiles and optical images, respectively. Three metadata files accompany these directories: sonar.csv – metadata for all SSS tiles. camera.csv – metadata for all optical images. correspondences.json – mapping between optical images and SSS tiles. The mapping metadata consists of dictionary objects, defined as unordered collections of values with unique keys. Each key represents the index of an SSS tile, and the corresponding value is a list of indices identifying the optical images associated with that tile based on the matching criteria. A summary of the SSS and camera metadata files is provided in Appendix B of the paper.

本数据集取自BenthiCat馆藏中的配准光学图像与侧扫声呐（side-scan sonar, SSS）样本，旨在支撑多模态学习研究。每幅光学图像均依据论文所述的地理邻近性原则，与对应的256×256像素侧扫声呐瓦片（SSS tile）建立空间关联。本数据集未包含任何图像级或像素级标注信息。数据集分为两个核心目录：sonar与camera，分别存储匹配得到的侧扫声呐瓦片与光学图像。配套提供三份元数据文件：sonar.csv——所有侧扫声呐瓦片的元数据；camera.csv——所有光学图像的元数据；correspondences.json——光学图像与侧扫声呐瓦片的映射关系表。该映射元数据采用字典对象格式，即由携带唯一键值的无序值集合构成。每个键代表一个侧扫声呐瓦片的索引，对应的值为一组索引列表，用于标识基于匹配准则与该瓦片关联的光学图像。侧扫声呐与光学图像元数据文件的详细摘要已收录于论文附录B中。