Paper 2508.14879
MeshCoder: LLM-Powered Structured Mesh Code Generation from Point Clouds
- Published
- Aug 2025
- Research lab
- Independent
- Citations
- 12
- GitHub
- 501 stars
01 In brief
Summary
MeshCoder is a framework that reconstructs 3D objects from point clouds into editable Blender Python scripts.
It introduces a set of expressive Blender Python APIs capable of modeling complex geometries beyond simple primitives, including primitives, translation, bridge loops, boolean operations, and arrays.
A large-scale paired object-code dataset was constructed using Infinigen Indoor, comprising 1 million objects across 41 categories, with each object decomposed into semantic parts.
A part-to-code inference model was trained on a synthetic dataset of 10 million part-code pairs, then used to assemble object-level codes.
A multimodal large language model (Llama-3.2-1B) with a triplane-based shape tokenizer was trained to generate executable code from point clouds.
Experiments show MeshCoder significantly outperforms baselines (Shape2Prog, PLAD) in reconstruction accuracy, achieving an average IoU of 86.75% and CD of 0.06×10^-2.
The code-based representation enables intuitive geometric and topological editing, and improves LLM reasoning in 3D shape understanding tasks.
Limitations include a focus on human-made objects, with organic forms left for future work.
02 From the paper
Abstract
Reconstructing 3D objects into editable programs is pivotal for applications like reverse engineering and shape editing. However, existing methods often rely on limited domain-specific languages (DSLs) and small-scale datasets, restricting their ability to model complex geometries and structures. To address these challenges, we introduce MeshCoder, a novel framework that reconstructs complex 3D objects from point clouds into editable Blender Python scripts. We develop a comprehensive set of expressive Blender Python APIs capable of synthesizing intricate geometries. Leveraging these APIs, we construct a large-scale paired object-code dataset, where the code for each object is decomposed into distinct semantic parts. Subsequently, we train a multimodal large language model (LLM) that translates 3D point cloud into executable Blender Python scripts. Our approach not only achieves superior performance in shape-to-code reconstruction tasks but also facilitates intuitive geometric and topological editing through convenient code modifications. Furthermore, our code-based representation enhances the reasoning capabilities of LLMs in 3D shape understanding tasks. Together, these contributions establish MeshCoder as a powerful and flexible solution for programmatic 3D shape reconstruction and understanding. The project homepage is available at \href{https://daibingquan.github.io/MeshCoder}{this link}.