The Science behind the Site

Welcome to our website! We would like to ask you to help us classify galaxies in Euclid.

Idea

The Euclid Wide survey will image one third of the sky (15,000 deg. squared), with a high angular resolution (better than 0.2 arcseconds). It will produce images of billions of objects, and we estimate that ~50 million galaxies will be sufficiently well resolved to determine their visual morphologies, whether they are spiral, elliptical or peculiar galaxies. For more than a decade, a standard way to classify galaxies has been through the visual inspection of images by volunteer citizen scientists. The sizes of modern surveys, such as Euclid, however, scale exponentially, which is not the case for the number of volunteers. New methods employing artificial intelligence, for example, convolutional neural networks, have been used to classify galaxies in an automated way. Nevertheless, these methods rely on large training sets in order to achieve an accurate classification of galaxies. These training sets have to be provided by humans. For this purpose, we designed this platform asking experts from the Euclid Consortium to classify Euclid-like images, in order to calibrate our automated galaxy classifiers and be ready for when Euclid will be launched.

Galaxies on this website

The classification of galaxy images depends on the resolution and depth of images. A galaxy imaged from the ground will appear more "blurry" compared to the same galaxy imaged from space. In order to be as close as possible to the quality of the images produced by Euclid, we use galaxies imaged by the Hubble Space Telescope (which has a better resolution than Euclid). In order to make galaxies look like they were imaged by Euclid, we convolved these images with the Euclid VIS PSF and added noise characteristic to Euclid. Furthermore, we added a set of emulated images from the Horizon-AGN simulations. These images are of galaxy mergers at various redshifts, which will help us identify and calibrate the stage of the mergers in Euclid.

Tests

We would like to calibrate our machine learning methods for the classification of Euclid galaxies by determining:

1. How many galaxies are needed in the training sets for the various features of galaxy morphology.
2. The effect of observational biases (redshift, magnitude, physical resolution) on the reliable classification of galaxies.

Science

Ultimately, your classifications will help us study the evolution of galaxy morphology for the last 9 billion years. We will be able to study the evolving trends, in sizes, structures, and morphologies, revealing the formation mechanisms behind galaxies and providing a new and unique way to test theories of galaxy formation. Some of the questions we would like to answer are:

How and when these structures form?
How do they shape the galaxies and affects their physical properties?
What is the role of the environment?
What is the role of secular evolution?
How did the first disc galaxies form and what transformations occurred in the history of the Universe?

With Euclid directly probe the dynamical properties of galaxies for a large fraction of the sky and cosmic history.