Oil Spill Detection in the Ocean Using U-net Convolutional Neural Network Architecture

Overview

This project presents an innovative approach to detecting oil in the ocean using Convolutional Neural Network (CNN) with U-net architecture. Developed in response to a significant oil spill in Ventanilla, Peru, our model employs satellite imagery for effective identification and analysis of marine oil spills.

Key Features

• Advanced Image Processing: Utilizing PeruSat-1 satellite imagery, processed for optimal neural network training.
• Customized CNN Models: Tailored U-net models for accurate hydrocarbon segmentation.
• Multiple Training Phases: Includes initial training, reinforced learning with additional datasets, and comparative analysis of pre-trained models.

Dependencies

• Python >= 3.8.10
• OpenCV
• NumPy
• Matplotlib
• TensorFlow

Repository Structure

• Architecture: Contains the architectures of trained models.
• DataSave: Stores trained models and training history.
• Dataset: SAR satellite image dataset.
• DatasetPeruSat1: PeruSat-1 satellite image dataset.
• main.ipynb: Primary Jupyter Notebook for project demonstration.

Getting Started

To run the main.ipynb notebook:

1. Ensure all dependencies are installed. You can install them using pip install -r requirements.txt if you have a requirements file.
2. Navigate to the folder containing main.ipynb.
3. Launch Jupyter Notebook or Jupyter Lab by running jupyter notebook or jupyter lab in your terminal or command prompt.
4. Open main.ipynb from the Jupyter interface.
5. Run the cells in sequence to observe the model training, evaluation, and visualization processes.

Note: The notebook is commented to guide you through each step of the process.

Trained Models

• SAR: Initial model trained with SAR satellite images.
• PER1: Model trained with PeruSat1 satellite images.
• SAR_PER1: SAR model enhanced with PeruSat1 dataset.
• PER1_SAR: PER1 model improved using SAR image dataset.

Performance and Evaluation

Models are rigorously evaluated using F1-Score metrics, ensuring high accuracy and reliability in hydrocarbon detection.