MeshCoder Dataset
收藏资源简介:
MeshCoder数据集是用于3D物体重建的大规模配对对象-代码数据集,包含约100万个物体,涵盖41个类别,每个物体最多可达100个部分。数据集通过使用Blender Python API合成复杂几何体,并训练多模态大型语言模型(LLM)将3D点云转换为可执行的Blender Python脚本。数据集旨在解决现有方法在建模复杂几何结构和缺乏大规模配对数据集的限制,用于3D形状重建和理解的程序化方法。
The MeshCoder dataset is a large-scale paired object-code dataset for 3D object reconstruction. It contains approximately 1 million objects spanning 41 categories, with each object having up to 100 parts. The dataset synthesizes complex geometries using the Blender Python API, and trains multimodal large language models (LLMs) to convert 3D point clouds into executable Blender Python scripts. This dataset aims to address the limitations of existing methods in modeling complex geometric structures and the lack of large-scale paired datasets, and is designed for procedural approaches to 3D shape reconstruction and understanding.




