Indoor-6

我们的室内6数据集是根据多天在六个室内场景中捕获的多个会话创建的。使用COLMAP计算每个场景的伪地面真相 (pGT) 3D点云和相机姿势。下图显示了相机的姿势 (红色) 和点云 (灰色)，并且对于每个场景，训练和测试中的视频和图像的数量分别分开。与7个场景相比，室内-6中的场景更大，具有多个房间，并且由于图像跨越多天和一天中的不同时间而包含照明变化。

Our Indoor-6 dataset is constructed from multiple sessions captured over multiple days across six indoor scenes. COLMAP was utilized to compute the pseudo-ground truth (pGT) 3D point clouds and camera poses for each scene. The figure below shows the camera poses (red) and point clouds (gray), and the quantities of videos and images in the training and test splits are provided separately for each scene. Compared with the 7-Scene dataset, the scenes in Indoor-6 are more expansive, featuring multiple rooms, and contain illumination variations since the images were collected across multiple days and different times throughout the day.

搜集汇总

背景与挑战

背景概述

Indoor-6是一个包含六个室内场景的多天捕获数据的数据集，提供伪地面真相3D点云和相机姿势，适用于计算机视觉和相机定位研究。