Skip to content

Dataset · YipengGao

Benchmark and data for procedural code-based 3D modeling

YipengGao/3DCodeBench

10,000 to 100,000 examples featuring text, images, and 3D code for evaluating and training 3D generation models.

What's inside

An open reproducible version of the 3DCodeBench benchmark accompanying a research paper on agentic procedural 3D modeling via code generation. It contains between 10K and 100K English elements combining text descriptions, images, and 3D data. The authors warn that code and data undergo final quality checks, so unexecutable scripts or mismatched renders descriptions are possible.

What you can do with it

  • Evaluating multimodal and code agents in text-to-3d and image-to-3d tasks
  • Training executable code generation models to build 3D geometry (after error filtering)
Tasks
Text → 3d, Image → 3d
Languages
English
Size
10K–100K rows
Downloads this month
300.5K
Likes
37

How to load it

from datasets import load_dataset

ds = load_dataset("YipengGao/3DCodeBench")