Common Objects in 3D 常见物体多视角图像数据集

Common Objects in 3D 是常见物体的多视角图像数据集。图像带有摄像机拍摄角度和 ground truth 3D 点云标注。该数据集包含近 19,000 个视频的共 150 万帧。这些视频捕捉了来自 50 个 MS-COCO 类别的物体。因此，它在类别和物体的数量上都优于其他数据集。

Common Objects in 3D is a multi-view image dataset of common real-world objects. The images are accompanied by camera pose annotations and ground-truth 3D point cloud labels. This dataset contains a total of 1.5 million frames across nearly 19,000 videos, which capture objects belonging to 50 MS-COCO categories. Consequently, it outperforms other competing datasets in terms of both the number of categories and the total number of objects.

搜集汇总

背景与挑战

背景概述

Common Objects in 3D 是一个多视角图像数据集，包含来自50个MS-COCO类别的近19,000个视频共150万帧图像，每帧均带有摄像机拍摄角度和真实3D点云标注。该数据集可用于大规模3D类别重构的学习与评估，其规模和类别数量均优于其他同类数据集。