VeRi
收藏资源简介:
VeRi数据集是一个用于车辆再识别的大型基准数据集,包含超过50,000张776辆车辆的图像,这些图像由20个摄像头在24小时内覆盖1.0平方公里的区域拍摄。数据集在真实世界的非约束监控场景中捕获,并标注了多种属性,如边界框、类型、颜色和品牌,支持复杂模型的学习和评估。每辆车由2到18个不同视角、光照、分辨率和遮挡的摄像头捕获,提供了高重复率,适用于实际监控环境中的车辆再识别。此外,数据集还标注了足够的牌照和时空信息,如牌照边界框、牌照字符串、车辆时间戳以及相邻摄像头之间的距离。
The VeRi dataset is a large-scale benchmark dataset for vehicle re-identification, comprising over 50,000 images of 776 vehicles captured by 20 cameras over a 24-hour period covering an area of 1.0 square kilometers. The dataset is captured in real-world, unconstrained surveillance scenarios and annotated with multiple attributes such as bounding boxes, type, color, and brand, supporting the learning and evaluation of complex models. Each vehicle is captured by 2 to 18 cameras with varying perspectives, lighting conditions, resolutions, and occlusions, providing a high repetition rate suitable for vehicle re-identification in practical surveillance environments. Additionally, the dataset is annotated with sufficient license plate and spatio-temporal information, including license plate bounding boxes, license plate strings, vehicle timestamps, and distances between adjacent cameras.
VeRi数据集概述
数据集特性
- 规模与覆盖范围:包含超过50,000张图像,涉及776辆车,由20个摄像头在1.0平方公里的区域内24小时内拍摄。
- 图像属性:图像在真实世界的非受控监控场景中捕捉,标注了多种属性,包括边界框(BBoxes)、类型、颜色和品牌。
- 车辆视角与环境:每辆车由2至18个不同视角、光照、分辨率和遮挡情况的摄像头捕捉,提供了高重复率。
- 标注信息:数据集还标注了足够的牌照和时空信息,如牌照框、牌照字符串、车辆时间戳以及相邻摄像头间的距离。
数据集下载
- 获取方式:数据集正在修订中,需通过电子邮件向联系人(xinchenliu at bupt dot edu dot cn)提供全名和单位信息以请求获取,确保数据用于非商业目的。
数据集问题
- 标注错误:先前版本的训练集标签文件(train_label.xml)中,车辆3、7和41被错误标注为类型ID="0",与合法类型列表不符。
引用信息
- 参考文献:使用此数据集时,请引用以下论文:
- Liu X., Liu W., Ma H., Fu H.: Large-scale vehicle re-identification in urban surveillance videos. In: IEEE International Conference on Multimedia and Expo. (2016) accepted.
- Liu X., Liu W., Mei T., Ma H. A Deep Learning-Based Approach to Progressive Vehicle Re-identification for Urban Surveillance. In: European Conference on Computer Vision. Springer International Publishing, 2016: 869-884.




