The Cityscapes dataset contains video sequences recorded in 50 different city street scenes, with 20,000 weakly annotated frames and 5,000 frames with high-quality pixel-level annotations. This dataset focuses on the semantic understanding of urban street scenes and aims to evaluate the use of visual algorithms in semantic urban scene understanding. The application of this dataset is as follows:

• Pixel-level and instance-level semantic labels;
• Research on large amounts of (weakly) annotated data. The Cityscapes dataset was released in 2016 by the Cityscapes team, which consists of people from Daimler AG, the Max Planck Institute for Informatics, the Visual Reasoning Laboratory of the Technical University of Darmstadt, etc. The related paper is "The Cityscapes Dataset for Semantic Urban Scene Understanding"