Using Tensorflow to Detect Objects in an Image

The abundance of various types of satellite imagery has created a challenge with processing and analyzing data into useful information. In this project, we explored the development, implementation, and evaluation of a machine learning algorithm, specifically a neural network, to automate the detection of ships to track traffic in a desired port or region. We also used a graphical approach to computation using TensorFlow, which offers easy massive parallelization and deployment to the cloud. The final result is an algorithm, which is capable of receiving images from various sources of imagery at various resolutions and be able to identify the appropriate objects within the image.

Keywords: machine learning, neural network, identification, ships, deep learning, image processing, TensorFlow