RailSem19 (RailSem19: A Dataset for Semantic Rail Scene Understanding)

RailSem19提供了从轨道车辆 (火车和电车) 的自我角度拍摄的8500独特图像。提供了广泛的语义注释，包括基于几何的 (与轨道相关的多边形，所有轨道均作为折线) 和具有许多城市景观兼容的道路标签的密集标签图。许多框架显示了公路和铁路车辆之间的交叉区域 (铁路道口，在城市街道上行驶的电车)。RailSem19适用于铁路应用和道路应用。

RailSem19 provides 8500 unique images captured from the first-person perspective of rail vehicles (trains and trams). It includes extensive semantic annotations, specifically dense label maps that contain geometry-based track-related polygon annotations (with all rails represented as polylines) and road labels compatible with various urban scenarios. Numerous frames depict crossover regions between road and rail vehicles, such as railway crossings and trams running on urban streets. RailSem19 is suitable for both railway and road applications.

搜集汇总

背景与挑战

背景概述

RailSem19是一个用于语义铁路场景理解的数据集，包含8500张从轨道车辆视角拍摄的图像，提供基于几何的语义注释和密集标签图，适用于铁路和道路应用，特别关注铁路道口等交叉区域。该数据集由奥地利理工学院于2022年发布，旨在支持自动驾驶和场景分析研究。