3DFRONT-NC
收藏资源简介:
本文介绍了3DFRONT-NC数据集,这是一个经过精细处理的数据集,旨在减少原始3DFRONT数据集中的物体交叉现象。该数据集通过将物体分解为立方体原语,使用自回归模型对场景进行生成,实现了紧凑的场景布局并有效避免了物体交叉。数据集的创建过程包括将3D网格进行体素化、粗粒化处理以及立方体合并,从而简化几何表示并减少离散元素。3DFRONT-NC数据集的应用领域主要在于室内场景合成,旨在解决虚拟现实、增强现实以及室内设计等领域中场景合成的挑战,提高生成场景的真实性和紧凑性。
This paper introduces the 3DFRONT-NC dataset, a meticulously processed dataset designed to mitigate object intersection issues in the original 3DFRONT dataset. This dataset decomposes objects into cube primitives and uses autoregressive models for scene generation, achieving compact scene layouts while effectively eliminating object intersections. The creation process of the 3DFRONT-NC dataset includes voxelization, coarse-grained processing, and cube merging of 3D meshes, which simplifies geometric representations and reduces discrete elements. The 3DFRONT-NC dataset is primarily applied to indoor scene synthesis, aiming to address the challenges of scene synthesis in fields such as virtual reality, augmented reality, and interior design, and enhance the realism and compactness of generated scenes.
CASAGPT: Cuboid Arrangement and Scene Assembly for Interior Design 数据集概述
数据集基本信息
- 名称: 3DFRONT-NC
- 相关论文: CASAGPT: Cuboid Arrangement and Scene Assembly for Interior Design (CVPR 2025)
- 状态: 准备中,即将发布
数据集内容
- 用途: 室内设计的立方体排列和场景组装
- 数据格式: 未明确说明(待发布)
获取方式
- 发布地址: https://github.com/CASAGPT/3DFRONT-NC
- 当前状态: 代码库整理中,数据集即将发布
相关资源
- 代码实现: PyTorch
- 代码库: 当前仓库(https://github.com/CASAGPT/CASA-GPT)为论文实现,不包含数据集




