Tanks Temple

Tanks Temple图像数据集提供高分辨率视频，研究人员可以从中收集图像并根据图像进行3D重建。数据集提供训练数据和测试数据两种类型，测试数据分为中间组和高级组。训练集提供7个场景的7个高分辨率视频。
测试集共提供14个场景的14个高分辨率视频: 中间组包含雕塑、大型车辆和带有外部摄像机轨迹的房屋建筑，高级组包含从内部和大型室外场景拍摄的室内场景，具有复杂的几何布局和摄像机轨迹。
该数据集由英特尔实验室的Arno Knapitsch、Jaesik Park、Qian-Yi Zhou和Vladlen Koltun 2017年在《纸罐和寺庙: 基准大规模场景重建》中发表。

The Tanks and Temples dataset offers high-resolution videos, enabling researchers to extract images for 3D reconstruction tasks. The dataset includes two types of data: training data and test data, where the test data is further divided into an intermediate group and an advanced group. The training set contains 7 high-resolution videos corresponding to 7 distinct scenes. The test set provides 14 high-resolution videos for 14 scenes in total: the intermediate group includes sculptures, large vehicles, and building structures captured with external camera trajectories, while the advanced group covers indoor scenes shot from both interior and large outdoor environments, featuring complex geometric layouts and camera trajectories. This dataset was published in 2017 by Arno Knapitsch, Jaesik Park, Qian-Yi Zhou, and Vladlen Koltun from Intel Labs in the paper titled *Tanks and Temples: Benchmarking Large-Scale Scene Reconstruction*.

搜集汇总

背景与挑战

背景概述

Tanks Temple是一个用于3D重建研究的高分辨率图像数据集，包含训练集（7个场景）和测试集（14个场景，分为中间组和高级组），覆盖雕塑、车辆、建筑等多种复杂场景。该数据集由Intel Labs于2017年发布，旨在支持大规模场景重建的基准测试。