LayeredDepth
收藏资源简介:
LayeredDepth数据集是由普林斯顿大学计算机科学系创建的,包含1500张真实世界场景图片和15.3万张合成图片。该数据集旨在支持多层深度估计任务,提供了高质量、人工注释的相对深度地面真实数据,是首个包含多层深度注释的数据集。数据集中的图片覆盖了家庭、餐馆、实验室、城市环境和艺术装置等多种场景,并在各种光照条件下拍摄。此外,数据集还包含一个合成数据生成器LayeredDepth-Syn,能够生成具有无限变化的对象和场景布局的合成数据。
The LayeredDepth Dataset was developed by the Department of Computer Science at Princeton University. It contains 1,500 real-world scene images and 153,000 synthetic images. This dataset is designed to support multi-layer depth estimation tasks, providing high-quality, manually annotated relative depth ground truth data, and is the first dataset to include multi-layer depth annotations. The images in the dataset cover various scenarios including homes, restaurants, laboratories, urban environments, and art installations, and were captured under diverse lighting conditions. Additionally, the dataset also includes a synthetic data generator named LayeredDepth-Syn, which can generate synthetic data with infinitely varied objects and scene layouts.




