Ship Detection and Analysis Project

Project Overview

This project focuses on detecting ships in satellite images as part of the Airbus Ship Detection Challenge on Kaggle. The goal is to identify and segment ships in satellite images, which is a crucial task for maritime safety, fishing regulation, and environmental protection. Due to time and resource constraints, the model was trained on 1% of the dataset provided in the challenge.

Requirements

• pandas
• numpy
• Pillow
• matplotlib
• tensorflow
• scikit-image

To install the required Python modules, run:

pip install -r requirements.txt

File Descriptions

• train.py: Python script for training the model on ship detection.
• test.py: Python script for evaluating the model on a test dataset.
• eda.ipynb: Jupyter notebook containing exploratory data analysis (EDA), visualizations, and initial insights into the dataset.

Setup Instructions

1. Ensure Python 3.6+ is installed on your system.
2. Install the required modules using the requirements.txt file.
3. Download and prepare your dataset in the specified directory structure. Due to resource limitations, training was conducted on 1% of the data.

Usage Guide

• Run train.py to start the training process. Customize parameters as needed.
• Use test.py to evaluate the trained model's performance on the test dataset.
• Explore eda.ipynb for a detailed walkthrough of data preparation, exploration, and initial insights.

Additional Notes

This project is designed for educational purposes and serves as a template for ship detection tasks. Adjust the model and parameters based on your specific dataset and requirements. The challenge and dataset can be explored further at the Airbus Ship Detection Challenge on Kaggle.