2026-10-14T12:43:38Z https://www.scidb.cn/oai

10.57760/sciencedb.40992 2026-06-26T15:59:17Z

2026-06-26 USMT(Underwater Sonar Moving Target) doi:10.57760/sciencedb.40992 en The USMT (Underwater Sonar Moving Target) dataset is the first publicly available low-pixel forward-looking sonar (FLS) image dataset for underwater moving targets worldwide. The data was collected in an artificial lake in Qing County, Cangzhou, Hebei Province, China, capturing a 533-mm-calibre, 6-meter-long Autonomous Underwater Vehicle (AUV). The dataset comprises 21 video sequences with a total of 7,136 fully annotated FLS images. Annotations are provided in LabelMe JSON standard for instance segmentation. The dataset includes both sector-shaped sonar images (632×792 pixels) and rectangular B-scan images (1,000×500 pixels). It is particularly suitable for developing small-target detection algorithms, investigating spatiotemporal fusion techniques for multi-frame target tracking, and benchmarking instance segmentation methods for underwater robotics applications. Forward-Looking Sonar; Underwater Moving Target; Instance Segmentation; Sonar Image; Low-Pixel Detection Jingfeng Yu Jiugen Lin Zhongju Sun Chen Liang PUBLIC https://creativecommons.org/licenses/by/4.0/ dataset Science Data Bank