UCM-VeID

该数据集名为UCM-VeID，是一个跨模态车辆重识别基准，包含了车辆的RGB和红外图像，旨在解决基于无人机进行车辆识别时数据不足的问题。该数据集包括16,015张RGB图像和13,913张红外图像，并标注了车辆类型、颜色和朝向。规模上，数据集涵盖了753个车辆身份，共29,928个视角标注，其任务是跨模态车辆重识别。

This dataset, named UCM-VeID, is a cross-modal vehicle re-identification benchmark containing RGB and infrared images of vehicles, aiming to address the issue of insufficient data for drone-based vehicle recognition. The dataset includes 16,015 RGB images and 13,913 infrared images, with annotations for vehicle type, color and orientation. In terms of scale, it covers 753 vehicle identities and has a total of 29,928 viewpoint annotations, and the core task of this benchmark is cross-modal vehicle re-identification.

搜集汇总

背景与挑战

背景概述

UCM-VeID是一个无人机跨模态车辆重识别基准数据集，包含两个版本（UCM-VeID和UCM-VeID_v2），用于支持相关研究。数据集通过百度网盘提供下载，并已在CVPR 2025等学术会议和期刊上发表相关研究成果。