OASIS (Open Annotations of Single Image Surfaces)

单视图3D是从单个图像中恢复3D属性 (例如深度和表面法线) 的任务。我们假设单图像3D的主要障碍是数据。我们通过呈现单个图像表面 (OASIS) 的开放注释来解决此问题，该单个图像3D的数据集由140,000图像的详细3D几何图形的注释组成。我们训练和评估各种单图像3D任务的领先模型。

Single-view 3D refers to the task of recovering 3D properties (e.g., depth and surface normals) from a single image. We assume that the primary obstacle to single-image 3D research is data. We address this issue by presenting Open Annotations for Surfaces of Single-image 3D (OASIS), a dataset consisting of detailed 3D geometry annotations for 140,000 images. We train and evaluate state-of-the-art models for various single-image 3D tasks.

搜集汇总

背景与挑战

背景概述

OASIS是一个用于单视图三维深度估计的大规模数据集，包含140,000张图像的详细3D几何注释。该数据集旨在通过提供丰富的3D数据来支持单图像3D任务的模型训练和评估。