DeepShipNet: Exploring neural network architectures for locating ships in images

This code is part of the final project of the University of California San Diego Fall Semester 2018 course Machine Learning for Image Processing. During the project, the team participated in the Kaggle Competition Airbus Ship Detection[1] using the same code.
Read the following instructions if you would like to run the code yourself.

Executable Files

For a quick demo, run the Jupyter Notebook Demo.ipynb.
For a full training on the dataset, run the Jupyter Notebook train.ipynb.

Requirements

The dataset can be downloaded on the kaggle competition website[1].
In the notebooks, the following locations are referenced.
Adapt the code or make sure to recreate the data structure:
csv data for labels: /datasets/ee285f-public/airbus_ship_detection/
training data: /datasets/ee285f-public/airbus_ship_detection/train_v2/
test data: /datasets/ee285f-public/airbus_ship_detection/test_v2/

To run the notebook, you need the following packages installed:

• tensorflow
• keras
• matplotlib
• pandas
• scikit.image
• scikit.learn

ECE 285 Group Project Members

Albrecht Wigand
University of California San Diego
[email protected]
U08106674

Jeffrey Wang
University of California San Diego
[email protected]
A11708879

Samuel Thornton
University of California San Diego
[email protected]
A53243732

Siddhant Jain
University of California San Diego
[email protected]
A53251799

Yan Sun
University of California San Diego
[email protected]
A53240727

References

[1] - https://www.kaggle.com/c/airbus-ship-detection