DUT-USEG
收藏资源简介:
DUT-USEG数据集由大连理工大学软件学院创建,是首个真实场景水下语义分割数据集,包含6617张水下图像,涉及海参、海胆、扇贝和海星四个类别。其中1487张图像具有语义分割和实例分割标注,剩余5130张具有目标检测框标注。数据集的创建旨在解决水下生物抓取技术中物体识别与分割的精度问题,特别是在提供物体轮廓等细致信息方面的需求。该数据集通过手工标注和基于已有目标检测数据集的扩展构建,适用于水下图像语义分割的研究和应用,特别是在提升水下机器人抓取效率和精度方面具有重要价值。
The DUT-USEG dataset was developed by the School of Software, Dalian University of Technology. It is the first real-world underwater semantic segmentation dataset, consisting of 6617 underwater images covering four categories: sea cucumber, sea urchin, scallop, and starfish. Among them, 1487 images are annotated with both semantic segmentation and instance segmentation ground truths, while the remaining 5130 images are only annotated with object detection bounding boxes. This dataset was created to address the accuracy challenges of object recognition and segmentation in underwater biological grasping technology, particularly to fulfill the demand for detailed information such as object contours. Constructed via manual annotation and expansion based on existing object detection datasets, the DUT-USEG dataset is applicable to research and applications of underwater image semantic segmentation, and holds significant value for improving the efficiency and accuracy of underwater robotic grasping.

- 1An Underwater Image Semantic Segmentation Method Focusing on Boundaries and a Real Underwater Scene Semantic Segmentation Dataset大连理工大学软件学院 · 2021年



