Cityscapes
收藏资源简介:
CitySeg/MOP数据集基于Cityscapes数据集,专门为实时语义分割设计,包含15个多目标优化问题。该数据集用于评估模型在精度、推理速度和硬件特定考虑等多目标上的表现。数据集的创建过程涉及将实时语义分割任务转换为标准的多目标优化问题,并通过EvoXBench平台提供与多种编程语言的无缝接口。CitySeg/MOP数据集主要应用于自动驾驶等实时应用场景,旨在解决模型设计中多目标优化的挑战。
The Cityscapes dataset was jointly created by Daimler AG, TU Darmstadt, MPI Informatics, and TU Dresden with the aim of advancing research in the visual understanding of complex urban street scenes. Renowned for its extensive scale, rich annotations, diverse scenarios, and high complexity, the dataset includes stereoscopic video sequences from 50 different city streets, with 5,000 images featuring high-quality pixel-level annotations; an additional 20,000 images are roughly annotated to support methods utilizing large amounts of weakly labeled data. Beyond pixel-level semantic annotations, the dataset also includes instance-level semantic annotations. To foster research on 3D scene understanding, depth information obtained through stereoscopic vision is provided as well. The dataset offers invaluable resources for research in fields such as autonomous driving, due to its unique complexity of urban internal traffic scenes.

- Cityscapes数据集首次发表,提供了高质量的图像分割数据,主要用于城市环境的语义理解研究。
- Cityscapes数据集首次应用于自动驾驶领域,成为该领域的重要基准数据集之一。
- Cityscapes数据集的扩展版本发布,增加了更多的图像和标注,提升了数据集的多样性和复杂性。
- Cityscapes数据集在计算机视觉顶级会议CVPR上被广泛讨论和引用,进一步巩固了其在学术界的影响力。
- Cityscapes数据集的应用扩展到智能交通系统,推动了城市交通管理的智能化发展。



