CARPK (car parking lot dataset)

停车场数据集 (CARPK) 包含通过无人机 (幻影3专业) 收集的来自4个不同停车场的近90,000辆汽车。在大约40米高度的无人机视图中收集图像。此外，我们从部分PUCPR数据集 (称为PUCPR数据集) 的单个图像中完成了所有汽车的注释，其中总共有近17,000辆汽车。这是一个大规模的数据集，用于在各种停车场的场景中进行汽车计数。图像集通过每辆车的边界框进行注释。所有标记的边界框都已被左上角和右下角的点很好地记录下来。它支持对象计数，对象定位以及在边界框中使用注释格式的进一步研究。

The Parking Dataset (CARPK) contains nearly 90,000 vehicles collected from 4 distinct parking lots via an unmanned aerial vehicle (Phantom 3 Professional). Images were captured at an altitude of approximately 40 meters from the drone's perspective. Additionally, we annotated all vehicles in individual images from a subset of the PUCPR Dataset, hereafter referred to as the PUCPR subset, with a total of nearly 17,000 vehicles. This is a large-scale dataset designed for vehicle counting tasks across diverse parking lot scenarios. The image set is annotated with bounding boxes for each individual vehicle, and all annotated bounding boxes are precisely recorded using their top-left and bottom-right corner coordinates. This dataset supports object counting, object localization, and further research utilizing the provided bounding box annotation format.

AI搜集汇总

背景与挑战

背景概述

CARPK是一个用于停车场场景的大规模汽车计数数据集，包含来自4个停车场的近90,000辆汽车图像，通过无人机在约40米高度采集，并补充了PUCPR数据集的约17,000辆汽车注释。所有图像均使用边界框标注，支持对象计数、定位等自动驾驶相关研究，由台湾大学和通用电气全球研究中心于2017年发布。