卡口车辆图像数据集（SYSU Cars）
本期开放车型精细识别数据集，开放的数据发布在OpenITS官方网站（http://www.openits.cn/），可供所有人免费下载使用，这些数据仅限用于非商业用途，且基于开放数据的研究成果必须标注数据由OpenITS提供。
1. 数据内容
卡口车辆图像数据集（SYSU Cars Dataset）是用于车型精细识别的实验室自建数据集，数据集示例如图1所示，数据集具体情况见参考论文：
李熙莹，全峰玮，叶芝桧. 基于混合类别均衡损失的车型精细识别[J/OL]. 计算机工程与应用，2022. https://kns.cnki.net/kcms/detail/11.2127.TP.20220713.1924.022.html.
图1 数据集示例
数据集包含102类品牌、691类型号、2516类年款，共66137张图片，其中包含49692张训练图片，16445张测试图片，训练集分布情况如图2所示。数据采集天数为连续2周，时间涵盖00:00:00-23:59:59。相较于 Stanford Cars 和 CompCars 平均每个品牌只包含 4 个型号，SYSU Cars平均每个品牌包含 6.77 个型号。由于相同品牌不同型号的车辆具有更高的相似性，因此 SYSU Cars具有更多难区分的样本，更具有挑战性。此外， SYSU Cars具有大量的车辆图片和丰富的车辆款式，更符合实际应用需求
图2 训练集分布情况
2. 数据文件
数据目录包含6个压缩包和1个csv文件：
sysu_cars_p1_11624.rar
sysu_cars_p2_12153.rar
sysu_cars_p3_12272.rar
sysu_cars_p4_6840.rar
sysu_cars_p5_12056.rar
sysu_cars_p6_11192.rar
SYSU Cars Label.csv
3. 数据文件描述
sysu_cars_p1_11624.rar：包含11624张jpg格式图片；
sysu_cars_p2_12153.rar：包含12153张jpg格式图片；
sysu_cars_p3_12272.rar：包含12272张jpg格式图片；
sysu_cars_p4_6840.rar：包含6840张jpg格式图片；
sysu_cars_p5_12056.rar：包含12056张jpg格式图片；
sysu_cars_p6_11192.rar：包含11192张jpg格式图片；
SYSU Cars Label.csv：包含每张图片对应标签。image_name：图片名称；pp_xh_nk：类别（品牌_型号_年款）；sub_nk：子年款类别；train_test_split：训练测试集划分（0-测试集、1-训练集）。
SYSU Cars Dataset
1. Dataset description
SYSU Cars Dataset is an image dataset used for fine-grained vehicle model recognition. A sample of data is shown in Fig.1. The details of the dataset are described in the paper:
LI Xiying，QUAN Fengwei，YE Zhihui. Fine-grained Vehicle Model Recognition based on Mixed Class Balance Loss[J/OL]. Computer Engineering and Applications，2022， https://kns.cnki.net/kcms/detail/11.2127.TP.20220713.1924.022.html
Fig.1 A sample of data
SYSU Cars contains 66137 images and there are 49692 images for training, 16445 images for testing. The distribution of training set is shown in Fig.2. Specifically, there are 102 brands, 691 models and 2516 items of year in the dataset. These images were collected from 00:00:00 to 23:59:59 in two weeks. Compared with Stanford Cars and CompCars, which contains 4 models per brand on average，the SYSU Cars contains 6.77 models per brand on average. While the images with same brand have higher similarity, the SYSU Cars contains more difficult samples and is more challenging. In addition, SYSU Cars contains a lot of vehicle images and models, which makes the dataset more useful.
Fig.2 The distribution of training set
2. Data file
The data contains six compressed packages and one csv files:
sysu_cars_p1_11624.rar
sysu_cars_p2_12153.rar
sysu_cars_p3_12272.rar
sysu_cars_p4_6840.rar
sysu_cars_p5_12056.rar
sysu_cars_p6_11192.rar
SYSU Cars Label.csv
3. Data file description
sysu_cars_p1_11624.rar：including 11624 images；
sysu_cars_p2_12153.rar：including 12153 images；
sysu_cars_p3_12272.rar：including 12272 images；
sysu_cars_p4_6840.rar：including 6840 images；
sysu_cars_p5_12056.rar：including 12056 images；
sysu_cars_p6_11192.rar：including 11192 images；
SYSU Cars Label.csv: including four columns of information, respectively corresponding to: image name (image_name), category (pp_xh_nk: Brand_Model_Year), Sub-year category (sub_nk) and image status (train_test_split: 0-test set, 1-train set).
数据提供单位
本数据由中山大学提供。
The dataset is provided by: Sun Yat-Sen University
引文格式（Citation Reference）：
中文引文格式如下，时间信息请按实际情况更改。
李熙莹，OpenITS联盟. OpenData V13.0-卡口车辆图像数据集（SYSU Cars）. http://www.openits.cn/openData4/822.jhtml (2022). Accessed: 2022-XX-XX.
李熙莹，全峰玮，叶芝桧. 基于混合类别均衡损失的车型精细识别[J/OL].. 计算机工程与应用，2022. https://kns.cnki.net/kcms/detail/11.2127.TP.20220713.1924.022.html.
Xiying Li, Fengwei Quan, Qianyin Jiang, Qiang Lu, "Structure-guided attention network for fine-grained vehicle model recognition," J. Electron. Imaging 31(2), 023033 (2022), doi: 10.1117/1.JEI.31.2.023033.
Please change the accessed data accordingly.
Li, X. & OpenITS. OpenData V13.0-SYSU Cars Dataset. http://www.openits.cn/openData4/822.jhtml (2022). Accessed: 2022-XX-XX.
LI Xiying，QUAN Fengwei，YE Zhihui. Fine-grained Vehicle Model Recognition based on Mixed Class Balance Loss[J/OL]. Computer Engineering and Applications，2022. https://kns.cnki.net/kcms/detail/11.2127.TP.20220713.1924.022.html
Xiying Li, Fengwei Quan, Qianyin Jiang, Qiang Lu, "Structure-guided attention network for fine-grained vehicle model recognition," J. Electron. Imaging 31(2), 023033 (2022), doi: 10.1117/1.JEI.31.2.023033.
相关联系人：
广东省智能交通系统重点实验室 吴荔非 openits@126.com
If you have any problems, please contacts：
Lifei Wu Sun Yat-Sen University, openits@126.com
数据集：
注 ：下载数据后解压时请使用除winRAR以外的解压工具进行解压