VehicleX, ObjectX, PersonX

本研究引入了三个新的合成数据集：VehicleX、ObjectX和PersonX，用于图像分类、人员重识别和车辆重识别任务。VehicleX包含1362个车辆身份，通过专业3D模型师手工制作的272个基础模型衍生而来，涵盖11种主流车辆类型。ObjectX和PersonX分别用于模拟分类数据和人员重识别数据，均具有丰富的外观多样性。这些数据集通过可编辑的对象和环境属性，在图形引擎中定义了可控的模拟环境，以生成大规模的训练集，旨在减少与真实世界数据的领域差距，特别是在内容层面。

This study introduces three novel synthetic datasets: VehicleX, ObjectX, and PersonX, tailored for image classification, person re-identification, and vehicle re-identification tasks respectively. VehicleX encompasses 1,362 vehicle identities derived from 272 base models manually created by professional 3D modelers, covering 11 mainstream vehicle categories. ObjectX and PersonX are respectively designed for simulating classification datasets and person re-identification datasets, both boasting rich visual appearance diversity. All these datasets establish controllable simulated environments in graphics engines via editable object and environmental attributes to generate large-scale training datasets, aiming to narrow the domain gap between synthetic and real-world data, especially at the content level.