XIMAGENET-12

XIMAGENET-12是由亚琛工业大学等机构的研究人员开发的一个视觉数据集，包含212,747张图像，旨在评估视觉模型在实际应用中的鲁棒性。数据集精选自ImageNet的12个类别，覆盖日常生活中的常见物体，并模拟了包括过度曝光、模糊和颜色变化等六种真实世界场景。通过精确的语义标注，该数据集能够深入分析背景对视觉模型的影响。此外，数据集还开发了一套量化标准来评估模型的鲁棒性，适用于研究视觉模型在复杂环境下的表现，如工业生产中的相机震动模拟。

XIMAGENET-12 is a visual dataset developed by researchers from RWTH Aachen University and other institutions. It contains 212,747 images and is designed to evaluate the robustness of visual models in real-world applications. Curated from 12 categories of ImageNet, the dataset covers common objects in daily life and simulates six real-world scenarios including overexposure, blur, and color variation. With precise semantic annotations, it enables in-depth analysis of the impact of background on visual models. Additionally, the dataset has established a set of quantitative evaluation criteria for assessing model robustness, which is suitable for researching the performance of visual models in complex environments, such as simulated camera shake in industrial production.