McWilliamsCenter / deep_galaxy_models

Repository hosting code for creating deep generative models for GalSim
10 stars 1 forks source link

Deep Generative Models for Galaxy Image Simulations

arXiv:2008.03833

This repository hosts the analysis code for Deep Generative Models for Galaxy Image Simulations, Lanusse, Mandelbaum, Ravanbakhsh, Li, Freeman, and Poczos (2020)

You can try out the generative model proposed in the paper with this live notebook from the GalSim-Hub repository: colab link

Content of this repository:

The data used to make the plots of the paper is available here: https://zenodo.org/record/3975700

How to use this repository

The first step in order to execute the notebooks is to download the data resulting from the main analysis script (see below for more details). You can do so using the following command:

$ wget -O results.tgz https://zenodo.org/record/3975700/files/results_lanusse2020.tar.gz?download=1
$ tar -xvzf results.tgz

This will download and extract the postage stamps of COSMOS images and of mock galaxy images used in the paper. The archive also contains catalogs of morphological statistics computed on each stamps.

With the data downloaded, the dependencies needed to run the notebooks are:

Then you should be able to run all the notebooks.