Parking Lot (PKLot)

该数据库包含在晴天，阴天和雨天从两个不同的停车场 (parking1和parking2) 捕获的12,417图像 (1280x720)。第一个停车场有两个不同的捕获角度 (parking1a和parking 1b)。图像被组织成三个目录 (parking1a、parking1b和parking2)。每个目录包含三个子目录，用于不同的天气条件 (多云，阴雨和晴天)。在每个子目录内，图像按采集日期组织。数据库的每个图像都有一个关联的XML文件，包括所有停车位的坐标及其标签 (占用/空置)。通过使用XML文件分割停车位，您将能够获得停车位的695,900图像。

This database contains 12,417 images (1280×720) captured under sunny, cloudy and rainy weather conditions from two distinct parking lots (parking1 and parking2). The first parking lot has two different capture angles (parking1a and parking1b). The images are organized into three directories: parking1a, parking1b and parking2. Each directory contains three subdirectories corresponding to different weather conditions: cloudy, rainy and sunny. Within each subdirectory, images are organized by their acquisition date. Each image in this database has an associated XML file that includes the coordinates of all parking spaces and their labels (occupied or vacant). By using these XML files to segment parking spaces, a total of 695,900 parking space images can be obtained.

搜集汇总

背景与挑战

背景概述

Parking Lot (PKLot) 是一个用于自动驾驶领域的停车位检测数据集，包含12,417张高分辨率图像（1280x720），覆盖晴天、阴天和雨天等多种天气条件，以及两个停车场的不同视角。数据集提供XML文件标注每个停车位的坐标和占用状态，通过分割可得到近70万张停车位图像，适用于计算机视觉任务如车位识别和状态分类。该数据集由学术机构于2015年发布，采用开放许可CC BY 4.0。