Galaxy Builder Results

For the past few years, a new Galaxy Zoo project has been under development. This project allows the creation of models of galaxies inside the Zooniverse website (although in a slightly experimental fashion). Many of you helped trial this project in December of 2017, and some have classified since it was quietly launched in late April. I’d like to take this opportunity to share with you some of the early results we have obtained from your classifications!

From the beta

260 of you helped beta test and debug the project in our beta, providing invaluable feedback (most of which we hope we addressed!) and submitting over one thousand test classifications of our small test sample. One of the images in this sample was the SDSS r-band image of the galaxy below, catchily known as SDSS J104238.12+235706.8.

In this post we’ll look specifically at the spiral arms drawn on this galaxy, and how we can recover information about the shape of the spirals from your drawn arms. If we plot every spiral arm drawn on the galaxy, two distinct spiral appear:

There are a number of classifications which have either crossed the center of the galaxy, linking the two arms with one line, or which attempt to enclose the spiral (as in the GZ:3D project) rather than tracing along the center line. This confusion arose from the short tutorial and confusing help text present in the beta, which was flagged by our testers (thank you!).

Grouping Lines

We can make use of unsupervised machine learning techniques to cluster together these drawn arms, and extract points corresponding to each arm (while throwing away some lines which didn’t fit into our groups).

Cleaning up

Taking one arm as an example, we identify points which could be considered outliers, and remove them to improve our later fit (using another unsupervised machine learning technique called Local Outlier Factor). In the image below, red dots correspond to points identified as outliers, and the blue contours can be seen as a probability map, with points in regions of darker blue being more likely to be an outlier.

Fitting a spiral

Brilliant – we now have a load of (unordered) points which roughly resemble the spiral arm of our galaxy! We’ll fit a smooth line to obtain a “best guess” of the galaxy’s spiral properties, the result of which can be seen below:


What we’ve discovered from the beta trial is that we need at least 20 attempts at drawing spirals on a galaxy to get reliable answers, so we’re keen to get more people trying to build galaxies on Galaxy Builder.

If you’d like to help please join us at If that’s not your thing we still need classifications on Galaxy Zoo: 3D, and the classic Galaxy Zoo is still live with all new images from the DECaLS survey.

About The Zooniverse

Online citizen science projects. The Zooniverse is doing real science online,.

One response to “Galaxy Builder Results”

  1. Jean Tate says :

    Great to see these results! 🙂

    Two questions:

    1) How – specifically – did your ML create the “best fit” spirals (i.e. the lines) from the selected sets of points?

    2) Have you tested the best fit spirals against any analytic form? In particular, how closely do these spirals match the canonical logarithmic spirals which seem ever-present in the literature?

    Looking forward to reading about other great results from this terrific project!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: