# PID-Symbol-Detection

**Repository Path**: dlml2/PID-Symbol-Detection

## Basic Information

- **Project Name**: PID-Symbol-Detection
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-09-09
- **Last Updated**: 2026-09-09

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# PID Symbol Detection using YOLOv11

### 📌 Project Overview
This repository contains an object detection project focused on Piping and Instrumentation Diagram (P&ID) symbol detection using the YOLOv11s model. The goal is to automatically identify and classify engineering symbols in P&ID diagrams, enabling faster digitalization, verification, and downstream integration for industrial automation and process engineering.

### 📊 Dataset

Size: ~12,000 images
Splits: Train / Validation / Test
Dataset was filtered and preprocessed to improve quality.
Each image contains multiple P&ID symbols annotated for detection.

### 🛠️ Methodology

Data Analysis: Inspected dataset distribution and symbol class representation.
Training: Used YOLOv11s for object detection. Initial training shows promising results but requires further fine-tuning for higher accuracy.
Evaluation: Currently evaluating model performance on test splits.

### 🚀 Results (Work in Progress)
Initial training demonstrates reasonable detection performance.
Accuracy can be improved with additional epochs, hyperparameter tuning, and possibly augmentation strategies.

<p align="center">
  <img src="images/result1.jpg" alt="Detection" />
</p>

### 🔮 Future Directions
This project is being extended into a P&ID LLM Helper, guided by another research paper. The idea is to:

1. Combine object detection results with LLM-based reasoning to automatically interpret entire diagrams.
2. Enable text-based querying of diagrams (e.g., "Which valves are connected to Pump P-101?").
3. Develop an AI assistant that bridges P&ID diagrams and natural language understanding for industrial applications.

### 📚 References
1. [YoLov5 for symbol extraction in PID Diagrams](https://www.researchgate.net/publication/366123842_YOLOv5_for_symbol_extraction_in_PID_diagrams)
2. [Paliwal et al., 2021](https://www.researchgate.net/publication/354461183_Digitize-PID_Automatic_Digitization_of_Piping_and_Instrumentation_Diagrams)

### 🙏 Acknowledgments
🔗 Special thanks to [ch-hristov](https://github.com/ch-hristov) for providing the dataset and releasing the associated paper that inspired this work. 
Community contributors and open-source maintainers of YOLOv11.