This project is solving the Neuromorphic Studies of Asteroid Imagery challenge. Description
DBNN.NASA is a project to classify asteroids based on light curves using neural networks. The DL4J library was used to train a deep neural network based on 6000 known asteroids with lightcurve data from the NASA WISE project.
Our program can automatically scrape the minor planet center’s website for notices of newly discovered asteroids, cross reference them with information from the WISE source database to extract light curve data and classify them with the neural network. The GUI consists of a table of the latest 50 asteroids, a map of the locations of the asteroids in earth-centric coordinates, and a visualization of the neural network in action. Users have the option to pull in fresh data off of the internet, or load asteroid references from a file.
In future, we would like to extend this project by running deep neural networks on WISE image data, as well as light curves, for automatic asteroid detection and classification.
License: Apache License 2.0 (Apache-2.0)
Source Code/Project URL: https://github.com/j3doucet/DBNN.NASA
Video - https://www.youtube.com/watch?v=_5a-Q4eLso4