Multi-sensor large-scale dataset for multi-view 3D reconstruction

本数据集名为‘多传感器大规模多视角3D重建数据集’，由斯科学院技术研究所主导创建。数据集包含约140万张图像，涉及107个场景，每个场景从100个视角在14种光照条件下进行拍摄。数据集主要用于评估和训练3D重建算法，特别是针对具有挑战性的材料属性，如无特征表面、高反射性及透明性。数据集的创建过程涉及使用多种传感器，包括智能手机、工业相机和结构光扫描仪，以确保数据的多样性和准确性。应用领域包括但不限于3D重建算法的评估和训练，以及相关任务的研究。

This dataset, officially named "Multi-Sensor Large-Scale Multi-View 3D Reconstruction Dataset", was developed and led by the Technical Institute of the Swedish Academy of Sciences. It contains approximately 1.4 million images spanning 107 distinct scenes, with each scene captured from 100 different viewpoints under 14 varying lighting conditions. Primarily designed for evaluating and training 3D reconstruction algorithms, this dataset specifically addresses challenging material properties such as featureless surfaces, highly reflective materials, and transparent objects. The dataset's development utilized a diverse array of sensors, including smartphones, industrial cameras, and structured light scanners, to ensure both the diversity and accuracy of the acquired data. Its application scope includes, but is not limited to, the evaluation and training of 3D reconstruction algorithms, as well as research on related tasks.