Ship Base Detection

• 

  Coast

• 

• 

  Small Ship

• 

  Goods Ship

• 

  White ship

Inspiration

We wanted to do object recognition in ML and wanted to contribute to the country's defence

What it does

It can take in images from drone or any aerial vehicle and can make out whether there is a ship or a fleet or ship or ship base from an image of land,sea or coast.

How we built it

• We used the Maritime satellite imagery dataset to train our neural network. It is dataset of about 5000 images of land, sea,coast,ships, ship base and so on.
• We converted those images into lower resolution images using our custom made neural network architecture in keras
• We split our dataset into training and testing dataset.
• We tweaked the neural network to achieve a 99% accuracy on the training and 96% on testing dataset
  

Challenges we ran into

• The images we got from the dataset had different resolutions so we first converted it into same resolution
• We made multiple changes to get a good enough neural network architecture like deciding on which layers to use, the number of layers, their depths,etc

Accomplishments that we're proud of

• We are proud of our 96% accuracy

What we learned

Making neural network from scratch

What's next for Ship Base Detection

• The neural network take into consideration the location of the image and the number of ships found and report unusual activity.
• We can add military establishment detection.
• We can add airforce establishment detection.

Built With

• anaconda
• keras
• opencv