AI-Based Isometric Diagram Extraction

AI-Based Isometric Diagram Extraction is a Python package for extracting meaningful structures from isometric diagrams using deep learning models. It leverages pre-trained models such as YOLO and anomaly detection autoencoders to detect, clean, and analyze diagrammatic representations.

Features

• Object Detection: Uses YOLO-based model (best.pt) to identify components in isometric diagrams.\
• Anomaly Detection: Uses autoencoder-based anomaly detection (anomaly.keras, my_model.keras).\
• Diagram Cleaning: Removes noise and enhances extracted diagrams.\
• Easy Inference: Run inference on any diagram image with a single function call.\
• Extensible: Future support for fine-tuning models on custom datasets.

Installation

You can install the package directly from source:

git clone https://github.com/nbeeeel/AI_Based_Isometric_Diagram_Extraction.git
cd AI_Based_Isometric_Diagram_Extraction
pip install .

Or after publishing to PyPI:

pip install ai-isometric-extractor

Usage

Python API

import ai_isometric_extractor as ai

# Run inference on an image
ai.run_inference("Test Pictures/334D92461.jpg")

CLI (coming soon)

ai-extractor input.jpg --output results/

Project Structure

AI_Based_Isometric_Diagram_Extraction/
│── ai_isometric_extractor/
│   ├── __init__.py
│   ├── main.py
│   ├── models/
│   │   ├── best.pt
│   │   ├── anomaly.keras
│   │   ├── my_model.keras
│── README.md
│── pyproject.toml
│── requirements.txt
│── LICENSE

Roadmap

• Initial package setup\
• Support for YOLO + Autoencoder models\
• CLI support\
• Fine-tuning on custom datasets\
• Extended visualization tools

License

This project is licensed under the MIT License - see the LICENSE file for details.

Author

Nabeel Ahmed
GitHub Profile