ai4ce/EUVS-Benchmark

该数据集包含104个城市场景，具有外推和内插的相机姿态。每个场景包含四个主要部分：图像、稀疏点云、训练集和测试集。该数据集适用于新视角合成、外推视角合成和3D重建等任务。This dataset comprises 104 urban scenes, featuring both extrapolated and interpolated camera poses. For each scene, it includes images, COLMAP format camera poses and sparse point clouds, training set and test set image names. The dataset is suitable for tasks such as novel view synthesis, extrapolated view synthesis, and 3D reconstruction.

搜集汇总

背景与挑战

背景概述

EUVS-Benchmark数据集包含104个城市场景，专注于外推和内推相机姿态，适用于新视角合成、外推视角合成和3D重建等计算机视觉任务。数据集提供图像、稀疏点云和相机姿态，总大小约25.6 GB，采用COLMAP格式，并包含明确的训练和测试集划分。