Sea-Ice Visual Perception and Risk Management Using Deep Neural Networks

Overview

This web-based application provides a two-stage deep learning image classification system to support sea-ice visual perception and risk management in polar navigation scenarios.

• Module I classifies the image perspective and conditions:

  • Forward Looking
  • Stern Looking
  • Side Looking
  • Lighting Condition
  • Irrelevant

• Module II further classifies valid images into:

  • Ice Images
  • Open Water
  • Objects

This tool was developed using Flask, YOLOv8, and deep neural network models trained on sea-ice image datasets.

---

📁 Project Structure

• project-root/

  • models/ (Trained model files)
  • static/ (Static assets - CSS, JS)
  • templates/ (HTML templates)
  • test_dataset/ (Sample test images)
  • prediction.py (Model prediction logic)
  • requirements.txt (Python dependencies)
  • server.py (Flask backend)
  • .gitignore (Git ignore rules)

  ---

🚀 Getting Started

1. Clone the Repository

git clone https://github.com/your-username/sea-ice-classification.git
cd sea-ice-classification

2. Set Up Environment

Make sure you have Python 3.8+ installed. You can use virtualenv:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

2. Install Dependencies

pip install -r requirements.txt

🖥️ Running the Application

Once dependencies are installed, launch the Flask app:

python server.py

By default, the app will be hosted on:

http://127.0.0.1:8080