Project motivation

  • This project aims to classify satellite ship images from google earth in real time.

Technicals

  • Seven types of ships were hand labelled and cleaned into 512x512 pixels, including boat, bulk cargo, container, cruiser, oil tank, submarine and yacht.
  • Ship images are sourced from MASATI v2 (MAritime SATellite Imagery dataset) and ShipRImageNet.
  • environment.yml file includes necessary packages to run training and inference.
  • Users can classify ships on google earth from browser, using the standalone module under folder "feed".
  • Interim presentation slides "ship classification interim.pdf" includes some examples of the ship images.
  • Final report "ship classification report.pdf" describes the tuning process in details.