Sunset Predictor

By Matthew Wilcox & Hanna Butsko

Background

This project uses Generative Adversarial Networks to build sky representations at Sunset. We built a GAN to just make sky images at sunset as well as a conditional GAN. The GAN was built based off of the DCGAN Pytorch Tutorial¹.

Data

We collected Image data from the NREL Solar Radiation Research Laboratory in Golden, Colorado ². Our Weather data was collected from Visual Crossing ³.

Example of the Processed Training Sunset Sky Image

GAN

Here is our Sunset Image GAN. It is built off of Random Noise to generate the images.

Example of the GAN Training process

Conditional GAN

Here is our Sunset Image Conditional GAN. It uses the weather conditions as predicted by the Sunset Images.

Example of the Conditional GAN Training

Presentation

Additionally, our presentation can be viewed here!

Footnotes

1. https://pytorch.org/tutorials/beginner/dcgan_faces_tutorial.html ↩

2. Andreas, A.; Stoffel, T.; (1981). NREL Solar Radiation Research Laboratory (SRRL): Baseline Measurement System (BMS); Golden, Colorado (Data); NREL Report No. DA-5500-56488. http://dx.doi.org/10.5439/1052221 ↩

3. https://www.visualcrossing.com/ ↩