🚆 Rail Defect Detection using YOLOv8

This project focuses on automating railway track defect detection using deep learning. Leveraging the YOLOv8-nano model and a user-friendly Streamlit interface, the system enables fast, real-time identification of defects in railway track images. It enhances railway safety and supports proactive maintenance.

---

📌 Overview

• Objective: Real-time defect detection in railway tracks from images.
• Core Tech: YOLOv8-nano (object detection), Streamlit (web interface).
• Dataset: Public datasets from Kaggle and Roboflow.
• Outcome: An accurate, scalable, and user-accessible tool for infrastructure monitoring.

---

🧠 Key Features

• 🔍 Real-Time Object Detection
  Detects railway defects with bounding boxes and confidence scores.

• 🖼️ Batch Image Upload & Processing
  Upload folders of images to process multiple images in one go.

• 📊 Accuracy Calculator
  Allows manual validation and model accuracy estimation.

• 🧾 Detection Summary Table
  View detected defect classes and confidence scores in a clear table format.

• 🧠 Caching for Reuse
  Avoids reprocessing already-checked images during the same session.

---

🛠 Tech Stack

Category	Technology
Model	YOLOv8-nano (Ultralytics)
Framework	PyTorch 2.6, CUDA 12.4
Interface	Streamlit
Dataset Source	Kaggle, Roboflow
Training Environment	Tesla T4 GPU
Language	Python

---

📂 Project Structure

rail-defect-detection/
├── data/
│   ├── train/
│   ├── val/
│   └── data.yaml
├── app/
│   └── defect_detection_app.py
├── models/
│   └── yolov8n.pt
├── notebooks/
│   └── model_training.ipynb
├── requirements.txt
└── README.md

---

🚀 Running the App Locally

1. Clone the Repository

   git clone https://github.com/yourusername/rail-defect-detection.git
   cd rail-defect-detection

2. Install Dependencies

   pip install -r requirements.txt

3. Start Streamlit App

   streamlit run app/defect_detection_app.py

---

📊 Model Evaluation

• Precision: Accurately identifies actual defects without false positives.
• Recall: Captures most actual defects with minimal false negatives.
• IoU (Intersection over Union): High bounding box accuracy.
• Manual Accuracy Tool: Users can validate and compare detection accuracy.

---

🎯 Future Enhancements

• ✅ Expand the dataset: Include more defect types like rust, wear, rail joint failures, etc.
• 📦 Optimize for embedded systems: Reduce model size and improve inference time for deployment on edge devices such as drones and robotic vehicles.
• 🔗 Integrate with maintenance alert systems: Enable auto-generation of work orders or notifications based on detected defects.
• 🎥 Video-based detection support: Add real-time video frame detection and annotation for continuous surveillance.

---

👨‍💻 Authors

• Pradeep Kumar
  B.Tech in AI & Data Science, Gati Shakti Vishwavidyalaya
• Harphool Singh
  B.Tech in AI & Data Science, Gati Shakti Vishwavidyalaya
• Asmit Sharma
  B.Tech in AI & Data Science, Gati Shakti Vishwavidyalaya
• **Vimal **
  M.Tech in Railway Engineering, Gati Shakti Vishwavidyalaya
• Mentor: Dr. Vipul Kumar Mishra
  Associate Professor, Department of AI & DS, Gati Shakti Vishwavidyalaya

---

📜 License

This project is intended for academic and research purposes only. For commercial or production use, please contact the project authors for permission.

---

📎 References

• Ultralytics YOLOv8 Documentation
• Railway Track Fault Detection Dataset – Kaggle
• Roboflow Rail Defect Dataset
• Streamlit Documentation