OpenMaterial

OpenMaterial是由洛桑联邦理工学院计算机视觉实验室和西安交通大学合作创建的大型数据集，专注于复杂材料的3D重建。该数据集包含1001个对象，涵盖295种不同材料，如导体、绝缘体、塑料及其粗糙变体，并在723种不同的光照条件下捕捉。数据集通过物理基础渲染技术生成高保真多视角图像，确保与真实世界对象的高度一致性。OpenMaterial提供全面的标注，包括3D形状、材料类型、相机姿态、深度和对象掩码，旨在为复杂材料对象的3D重建算法提供量化评估，推动相关技术的发展。

OpenMaterial is a large-scale dataset created through the collaboration between the Computer Vision Laboratory of École Polytechnique Fédérale de Lausanne and Xi'an Jiaotong University, focusing on 3D reconstruction of complex materials. This dataset includes 1001 objects covering 295 different material types, such as conductors, insulators, plastics and their rough variants, and was captured under 723 distinct lighting conditions. High-fidelity multi-view images are generated for the dataset via physics-based rendering technologies, ensuring a high degree of consistency with real-world objects. OpenMaterial provides comprehensive annotations including 3D shapes, material types, camera poses, depth information and object masks, aiming to provide quantitative evaluation for 3D reconstruction algorithms of complex material objects and promote the development of related technologies.