SeeThrough3D synthetic dataset
收藏资源简介:
SeeThrough3D合成数据集由印度海得拉巴国际信息技术学院与印度科学学院联合构建,旨在解决文本到图像生成中的3D遮挡推理问题。该数据集通过虚拟环境布置多样化的3D资产,并渲染包含强遮挡关系的多视角场景图像,数据内容涵盖复杂物体布局与相机参数控制。其创建过程采用Blender进行场景合成与渲染,重点模拟真实世界中的物体遮挡效应。该数据集支撑了遮挡感知的3D场景生成研究,可应用于游戏设计、建筑可视化等需要精确几何一致性的领域。
The SeeThrough3D synthetic dataset was jointly developed by the International Institute of Information Technology Hyderabad and the Indian Institute of Science, aiming to address the 3D occlusion reasoning problem in text-to-image generation. This dataset arranges diverse 3D assets in virtual environments, and renders multi-view scene images featuring strong occlusion relationships. The dataset encompasses complex object layouts and allows control over camera parameters. The creation process uses Blender for scene synthesis and rendering, with a focus on simulating real-world object occlusion effects. This dataset supports occlusion-aware 3D scene generation research, and can be applied to fields requiring precise geometric consistency such as game design and architectural visualization.



