---

Overview

Arti-PG is a procedural generation toolbox designed to efficiently synthesize large-scale and diverse 3D articulated objects with rich annotations. It models each object through a generalized structural program, providing analytic correspondence between parametric descriptions and geometric representations. By manipulating these structure programs via procedural rules, Arti-PG can produce unlimited shape variations and automatically attach semantic and affordance-level annotations. Arti-PG toolbox supports 26 categories of articulated objects, enabling comprehensive data generation for both vision and manipulation research tasks.

---

Environment Setup

# Create environment
conda create -n artipg python=3.9
conda activate artipg

# Install dependencies
pip install -r requirements.txt

---

Usage

Generate Objects and Parameters

Run the following command from the repository root:

python generate_objects.py --category Bottle --output ./outputs --gen_num 5

Arguments:

• --category: Name of the category (e.g., Bottle, Kettle, etc.)

• --output: Path to save the generated files

• --gen_num: Number of objects to generate

Each run will produce a series of:

• .obj files → Generated 3D mesh models

• .json files → Parameters describing the synthesized structure

Example Output:

outputs/
└─ Bottle/
   ├─ Bottle_000.obj
   ├─ Bottle_000.json
   ├─ Bottle_001.obj
   └─ Bottle_001.json

---

Digital Gene Project

The Digital Gene project seeks to understand the physical world through analytic concepts. While Arti-PG focuses on procedurally generating articulated 3D objects as a means of studying the physical world, it can also output the underlying parametric descriptions used in the generation process.

These parameters can be naturally interpreted using the analytic concepts provided by Digital Gene, forming a bridge between geometric synthesis and physical world understanding. To explore how the generated object parameters relate to Digital Gene theory and applications, please visit our companion repository:

👉 DigitalGene Project

---

Citation

If you find this work useful for your research, please cite our paper:

@inproceedings{sun2025arti,
  title={Arti-pg: A toolbox for procedurally synthesizing large-scale and diverse articulated objects with rich annotations},
  author={Sun, Jianhua and Li, Yuxuan and Wei, Jiude and Xu, Longfei and Wang, Nange and Zhang, Yining and Lu, Cewu},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={6396--6405},
  year={2025}
}