全天候道路图像分割数据集UAS(UESTC All-Day Scenery)

UAS共包含6380张不同时间和天气条件的图像，包括1399张在黄昏拍摄的样本、2167张在夜间拍摄的样本，819张在雨中拍摄的样本和1995张在阳光下拍摄的样本。对于每个图像，通过手动注释来创建精确的二进制语义标签。 整个数据集包含四种天候(黄昏天候，夜间天候，下雨天候和艳阳天候)共计6380张图像。其中，我们1使用海康威视DS-2CD2155F(D)-I(W)S采集黄昏天候，夜间天候，艳阳天候的道路图像；使用HUAWEI Mate 8采集下雨天候的道路图像。更详细的信息请见下表： 天候类型 数量 黄昏 1399（训练集1243，测试集156） 夜间 2167（训练集1926，测试集241） 雨天 819（训练集728，测试集91） 艳阳 1995（训练集1773，测试集222） 对于每一张采集的图像，UAS提供一张单通道类别标签和一张三通道的标记示意图。在类别标签中，0表示可行区域，而1表示不可行区域。采集图像与标记示意图如下图所示： 黄昏天候： 夜间天候： 下雨天候： 艳阳天候： @article{zhang2018road, title={Road segmentation for all-day outdoor robot navigation}, author={Zhang, Yuxiao and Chen, Haiqiang and He, Yiran and Ye, Mao and Cai, Xi and Zhang, Dan}, journal={Neurocomputing}, volume={314}, pages={316--325}, year={2018}, publisher={Elsevier} }

UAS contains a total of 6380 images captured under varying time and weather conditions, including 1399 twilight-shot samples, 2167 nighttime-shot samples, 819 rainy-day samples, and 1995 sunny-day samples. Precise binary semantic labels are created via manual annotation for each image. The entire dataset consists of 6380 images across four weather conditions: twilight, nighttime, rainy, and sunny. Specifically, road images under twilight, nighttime and sunny conditions were collected using a Hikvision DS-2CD2155F(D)-I(W)S camera, while road images under rainy conditions were captured with a HUAWEI Mate 8 smartphone. For more detailed information, please refer to the following table:
| Weather Condition | Total Count | Training Set Size | Testing Set Size |
|-------------------|-------------|-------------------|------------------|
| Twilight | 1399 | 1243 | 156 |
| Nighttime | 2167 | 1926 | 241 |
| Rainy | 819 | 728 | 91 |
| Sunny | 1995 | 1773 | 222 |
For each captured image, UAS provides a single-channel category label and a three-channel marking schematic diagram. In the category label, 0 represents the drivable area, while 1 represents the non-drivable area. The captured images and their corresponding marking schematic diagrams are shown below:
Twilight weather condition:
Nighttime weather condition:
Rainy weather condition:
Sunny weather condition:

@article{zhang2018road, title={Road segmentation for all-day outdoor robot navigation}, author={Zhang, Yuxiao and Chen, Haiqiang and He, Yiran and Ye, Mao and Cai, Xi and Zhang, Dan}, journal={Neurocomputing}, volume={314}, pages={316--325}, year={2018}, publisher={Elsevier} }

搜集汇总

背景与挑战

背景概述

UAS数据集包含6380张不同时间和天气条件的道路图像，分为黄昏、夜间、雨天和艳阳四种类型，每张图像配有精确的二进制语义标签和标记示意图，适用于全天候环境下的道路分割研究。