3D_Gen

3D Model Generation with RAG

Generate 3D printable models from text descriptions using DeepSeek-V3 enhanced with Retrieval-Augmented Generation (RAG).

Quick Start

1. Edit Prompts: Open 3D_Gen.py and modify the shape description prompt
2. Run the Generator: python 3D_Gen.py
3. Get Results: STL files are saved in the output directory

# Example shape generation
python 3D_Gen.py --shape "hollow cylinder with radius 15mm, height 40mm, wall thickness 2mm"

Features

• Generate 3D printable STL files from text descriptions
• Support for hollow shapes with precise dimensions
• Built-in validation for printability
• Measurement tools for dimensional accuracy

Repository Structure

• 3D_Gen.py - Main generation script (edit this to create your models)
• stl_analyzer.py - Tool for measuring generated models
• run_evaluation.py - Full evaluation system used in the research
• evaluation_results_date/ - Folder containing models and results generated by the system from our two main evaluations

Advanced Usage

For researchers interested in evaluating RAG performance for 3D model generation, see the detailed RAG Evaluation README which documents the experimental setup used in our research.

Requirements

• Python 3.9+
• Trimesh
• numpy-stl
• DeepSeek API key (for code generation)

pip install openai requests beautifulsoup4 sentence-transformers faiss-cpu trimesh numpy numpy-stl pyglet manifold3d shapely pandas matplotlib seaborn