CityScapes

Cityscapes是一个大型数据库，专注于对城市街道场景的语义理解。它为分为8个类别 (平面，人类，车辆，构造，对象，自然，天空和虚空) 的30个类提供语义，实例和密集的像素注释。数据集由大约5000个精细注释图像和20000个粗糙注释图像组成。在几个月，白天和良好的天气条件下，在50个城市中捕获了数据。它最初被记录为视频，因此手动选择帧以具有以下功能: 大量动态对象，不同的场景布局和不同的背景。Cityscapes is a large-scale database focused on semantic understanding of urban street scenes. It provides semantic, instance-level, and dense pixel-level annotations for 30 fine-grained categories grouped into 8 super-categories: flat surfaces, humans, vehicles, constructions, objects, nature, sky, and void. The dataset comprises approximately 5,000 finely annotated images and 20,000 coarsely annotated images. The data was collected across 50 cities over several months, under daytime and favorable weather conditions. It was originally recorded as video footage, so frames were manually selected to feature abundant dynamic objects, diverse scene layouts, and varied backgrounds.

搜集汇总

背景与挑战

背景概述

CityScapes是一个专注于城市街道场景语义理解的大型数据集，提供约5000个精细注释和20000个粗糙注释图像，涵盖30个类别（如人类、车辆、自然等），数据在50个城市的不同条件下捕获，适用于计算机视觉中的语义分割任务。