Digital Twin Tracking Dataset v2 (DTTD2)
收藏资源简介:
Digital Twin Tracking Dataset v2 (DTTD2)是由加州大学伯克利分校的研究团队开发的一个专注于数字双胞胎对象跟踪场景的RGBD数据集。该数据集扩展自DTTD v1,通过使用先进的移动RGBD传感器套件在Apple iPhone 14 Pro上捕获数据,增加了对iPhone传感器数据的适用性。DTTD2包含18个刚性对象及其纹理3D模型,数据来自100个场景,每个场景中包含一个或多个对象在各种方向和遮挡下的情况。数据集提供了3D对象姿态和每像素语义分割的地面实况标签,以及详细的相机规格、针孔相机投影矩阵和畸变系数。DTTD2旨在解决移动AR应用中动态环境下的3D对象定位问题,通过提供高质量的标注和多样化的场景,增强了数据驱动对象姿态估计算法的鲁棒性和实用性。
Developed by the research team from the University of California, Berkeley, Digital Twin Tracking Dataset v2 (DTTD2) is an RGBD dataset focused on digital twin object tracking scenarios. Derived from DTTD v1, this dataset expands its applicability to iPhone sensor data by capturing data using a state-of-the-art mobile RGBD sensor suite on an Apple iPhone 14 Pro. DTTD2 includes 18 rigid objects paired with their textured 3D models, collected across 100 scenes, each of which contains one or more objects under diverse orientations and occlusion conditions. The dataset provides ground-truth labels for 3D object poses and per-pixel semantic segmentation, alongside detailed camera specifications, pinhole camera projection matrices, and distortion coefficients. DTTD2 aims to address the problem of 3D object localization in dynamic environments for mobile AR applications, and enhances the robustness and practicality of data-driven object pose estimation algorithms by providing high-quality annotations and diverse scene configurations.




