RGB-D-D

RGB-D-D数据集是由北京交通大学信息科学研究所创建的大型深度图超分辨率基准数据集，包含4811对从室内到户外挑战场景的低分辨率(LR)和高分辨率(HR)深度图。该数据集通过移动电话和Lucid Helios相机分别捕捉LR和HR深度图，旨在解决现有方法在真实世界深度图SR任务中的局限性。数据集不仅适用于深度图SR研究，还可广泛应用于移动电话和其他深度相关任务，如人像摄影、物体建模、深度估计等。创建过程中，研究团队采用了多种技术和方法确保数据质量，如使用过分割算法和颜色化方法处理深度图，以及通过用户研究评估填充的HR深度图质量。RGB-D-D数据集的建立填补了理论研究与实际应用之间的空白，为深度相关任务提供了新的基准数据集。

The RGB-D-D dataset is a large-scale depth map super-resolution (SR) benchmark dataset developed by the Institute of Information Science, Beijing Jiaotong University. It comprises 4811 pairs of low-resolution (LR) and high-resolution (HR) depth maps spanning challenging scenarios from indoor to outdoor environments. LR and HR depth maps were captured using mobile phones and Lucid Helios cameras respectively, with the goal of addressing the limitations of existing methods in real-world depth map SR tasks. This dataset is not only applicable to depth map SR research, but also can be widely employed in applications related to mobile phones and other depth-centric tasks, such as portrait photography, object modeling, depth estimation and more. During the dataset development process, the research team adopted multiple technical approaches to ensure data quality, including applying over-segmentation algorithms and colorization methods to process depth maps, as well as evaluating the quality of the filled HR depth maps via user studies. The establishment of the RGB-D-D dataset bridges the gap between theoretical research and practical applications, providing a new benchmark dataset for depth-related tasks.