CCPD-Green

CCPD从中国一个省会城市的所有街道的路边停车场收集数据，那里的残余物拥有数百万辆汽车。每个停车收费员 (PFC) 每天从上午07:30到下午10:00在一条街道上工作，无论天气如何。对于每个停车帐单，收集者都需要使用Android手持pos机为汽车拍照，并手动注释确切的LP编号。值得注意的是，由于手持设备的不确定位置和拍摄角度，以及不同时间和不同街道上的不同照明和不同背景，来自手持设备的图像表现出强烈的变化。CCPD中的每个图像在有关LP的几个方面都有详细的注释 :( i) LP编号。(ii) LP边界框。(iii) 四个顶点位置。(iv) 水平倾斜程度和垂直倾斜程度。(v) 其他相关信息，例如LP区域，亮度，模糊度等。

CCPD is collected from roadside parking lots across all streets in a provincial capital city in China, which is home to millions of motor vehicles. Each parking fee collector (PFC) works on a designated street from 7:30 AM to 10:00 PM daily, regardless of weather conditions. For each parking ticket, the collector needs to take a photo of the vehicle using an Android handheld POS machine and manually annotate the exact license plate (LP) number. Notably, due to the uncertain positioning and shooting angles of the handheld device, as well as varying lighting conditions and backgrounds at different times and across different streets, the images captured by such devices exhibit significant variations. Every image in CCPD is comprehensively annotated with several aspects related to license plates: (i) LP number; (ii) LP bounding box; (iii) four vertex coordinates; (iv) horizontal tilt degree and vertical tilt degree; (v) other relevant information such as LP area, brightness, blurriness, etc.

搜集汇总

背景与挑战

背景概述

CCPD-Green是一个包含中国城市路边停车场车辆牌照图像的数据集，具有详细的牌照注释信息，适用于计算机视觉和图像识别研究。数据集由中国科学技术大学于2018年发布，图像因不同拍摄条件而具有较大变化性。