xorastro
Extremely Open, Extremely Reproducible Astronomy on deep learning on galaxy images for colors
The basic idea of this hack is to test out doing extremely open science (or live science) and extremely reproducible science with a small deep learning project.
We have been taking video ourselves throughout the hack and posting it to twitter as we do it. We'll save all that video, along with our slack channel to permanent resources.
The project is an attempt to measure galaxy (g-i) colors based only on their r-band image data, using an architecture similar to Dieleman et al 2015.