ELAS

该数据集是在巴西使用GoPRO HERO 3相机在不同的日子以每秒29.97帧的速度记录的。它包括22个场景 (超过15,000帧)，其中包含我们期望在现实世界中遇到的各种挑战: 高速公路，城市道路，交通，交叉路口，车道变更，道路上的文字，阴影，抖动捕获，不同的天气，略有不同的相机位置，褪色的车道标记，多个LMT转换，不同的车速等。在三个城市记录了场景: 巴西圣埃斯皮里图州的维托里亚 (vit ó ria)，维拉维拉 (Vila Velha) 和瓜拉帕里 (Guarapari)。每帧有640x480像素。该数据集已为研究界的相关事件手动注释，例如: 车道估计，更改和居中; 道路标记; 交叉路口; LMTs; 人行横道; 和相邻车道。

This dataset was recorded using a GoPRO HERO 3 camera in Brazil at 29.97 frames per second across multiple days. It comprises 22 scenes (over 15,000 frames) that encompass various real-world challenges, including highways, urban roads, traffic scenarios, intersections, lane changes, text on roads, shadows, jittery captures, diverse weather conditions, slightly varying camera positions, faded lane markings, multiple LMT transitions, different vehicle speeds, and more. The scenes were captured in three cities: Vitória, Vila Velha, and Guarapari, all located in Espírito Santo, Brazil. Each frame has a resolution of 640×480 pixels. This dataset has been manually annotated for relevant events for the research community, such as lane estimation, lane changes and centering, road markings, intersections, LMTs, crosswalks, and adjacent lanes.

搜集汇总

背景与挑战

背景概述

ELAS数据集是在巴西三个城市使用GoPRO相机采集的驾驶场景视频数据，包含22个场景超过15000帧，涵盖高速公路、城市道路等多种条件，并已手动标注车道、道路标记等事件。该数据集由Universidade Federal do Espírito Santo于2018年发布，旨在支持车道估计等相关研究。