Envisioning Distant Worlds
October 2023
This project explores the capabilities of GenerativeAI in predicting and visualizing distant planets. These planets, known as exoplanets, exist in our universe as points of data; too far away to witness with today’s technology.
This project utilizes principles in astrophysics, geophysics, chemistry, and mathematics to predict, and a fine-tuned version of Stable Diffusion 1.5 to visualize an exoplanet given its numeric data acquired via NASA telescopes. The results below are the first generated image for the input prompt into the trained model. The images have not been edited in any way.
Immediately following this text you will find the images for exoplanet HD 106315b and HD 56957b, respectively.
This project was conducted as a Master’s thesis for the University of Arts London, which was received with Distinction. This research was also presented as a stand-out paper during NeurIPS in December 2023.
The below images depict exoplanets KELT-9b and HD 4308b, respectively.
To read the research paper and code for this project, visit here.
Further information is available via my personal website at marissabeaty.me.