3D Object Renders

该数据集包含了五个不同3D对象的多种角度渲染图像，旨在评估通用扰动方法对抗敌意攻击的有效性。这些图像以224x224的分辨率渲染，并包含了摄像头角度变化和自然物体朝向的多样性，以模拟真实世界情境。数据集的规模为每个3D对象从十个不同角度进行渲染，其任务是对3D对象识别中的敌意攻击进行评估。

This dataset contains multi-angle rendered images of five distinct 3D objects, which is designed to evaluate the efficacy of universal perturbation methods against adversarial attacks. All images are rendered at a resolution of 224×224, and feature diverse camera angles and natural object orientations to mimic real-world scenarios. Each 3D object is rendered from ten distinct angles, and the core task of this dataset is to assess adversarial attacks in 3D object recognition.