Stereo Image Dataset (SID)
收藏资源简介:
Stereo Image Dataset (SID) 是由密歇根大学迪尔伯恩分校电气与计算机工程系创建的大型立体图像数据集,专门用于自动驾驶在恶劣条件下的研究。该数据集包含27个序列,总计超过178k立体图像对,涵盖从晴朗天空到夜间大雪等多种天气和光照条件。数据集的创建过程使用了ZED立体相机,记录了详细的天气、时间和道路条件注释,以及相机镜头污染的实例。SID旨在支持高级感知算法的开发和测试,特别是在现有数据集中未充分代表的条件下,如雪和雨。该数据集的应用领域包括自动驾驶车辆和高级驾驶辅助系统,旨在提高这些系统在各种天气和光照条件下的可靠性和一致性。
Stereo Image Dataset (SID) is a large-scale stereo image dataset created by the Department of Electrical and Computer Engineering, University of Michigan-Dearborn, specifically for autonomous driving research under harsh conditions. It consists of 27 sequences with a total of over 178k stereo image pairs, covering diverse weather and illumination conditions ranging from clear skies to heavy snow at night. The dataset was captured using a ZED stereo camera, with detailed annotations including weather, time, road conditions, as well as instances of camera lens contamination. SID aims to support the development and testing of advanced perception algorithms, especially in conditions that are underrepresented in existing datasets, such as snow and rain. Its application fields include autonomous vehicles and Advanced Driver-Assistance Systems (ADAS), with the goal of improving the reliability and consistency of these systems across various weather and illumination conditions.
数据集概述
数据集名称
SID: Stereo Image Dataset for Autonomous Driving in Adverse Conditions
数据集描述
SID数据集是为了支持自动驾驶系统在恶劣天气和光照条件下的高级研究而精心策划的。该数据集包含超过178k对高分辨率立体图像,分为27个序列,反映了雪、雨、雾和低光等多种条件。它涵盖了驾驶场景和环境背景的动态变化,包括大学校园、住宅街道和城市设置。该数据集旨在通过部分遮挡的摄像头镜头和可见度变化等场景,挑战感知算法,促进稳健的计算机视觉模型的发展。图像数据以标准PNG格式提供,无需专用软件或脚本即可访问。然而,研究人员和开发者可能需要他们的图像处理和计算机视觉工具包来有效利用该数据集。
关键词
- 自动驾驶
- 恶劣天气
- 立体视觉
- 图像数据集
- 计算机视觉
- 感知算法
创建者
- El Shair, Zaid A.
- Abu-raddaha, Abdalmalek
- Cofield, Aaron
- Alawneh, Hisham
- Aladem, Mohamed
- Hamzeh, Yazan
- Rawashdeh, Samir A.
数据集链接
数据集标识符
doi:10.7302/esz6-nv83
许可证
Attribution 4.0 International (CC BY 4.0)
发布机构
University of Michigan - Deep Blue Data




