FAQ
What is an artifact in astronomical imaging?
An artifact is a feature in an image that appears from effects of the camera, the optics, the detector, electronics, or other unknown reasons. The astronomical objects that the image was meant to capture, such as the spectra of galaxies, are not artifacts.
Why do we want to identify artifacts?
Identifying artifacts helps astronomers separate out real data, and understand the way the telescope instruments may be affecting that data. The results from this project will be used for both Euclid and in preparation for NASA's Nancy Grace Roman Space Telescope.
An image has both a spectrum and an artifact. What should I classify it as?
Choose the "Combination of objects" button.
My image has no obvious source in the green box. What do I do?
Click the "No obvious subject or other" button.
When do I use the Before Continuum Subtraction Images?
These are to help determine if a feature is really an artifact or not. If the object highlighted by the green box is a horizontal white stripe in the Before image, it is a spectrum, not an artifact.
How is it possible for a small blob not over a spectral continuum stripe to be an emission line?
If a galaxy is faint, the continuum will be very dim. Since the emission lines are brighter than the average light in the spectrum, sometimes only these bright peaks can be seen. We need to keep any small blob only a few pixels wide as data to analyze later, since it could be an emission line from a dim galaxy.
What are the black and white stripes?
These are the leftover pixels from continuum subtraction. We do our best to make the subtraction smooth, but sometimes there are remnants. These are not artifacts.
What are the solid black pixels?
In the before continuum subtraction image, you may notice lots of black pixels. We might consider these pixels “bad” due to a dead pixel, a signal from the internal heat of the camera, a defect in the chip, or something else. In order to ignore these pixels during artifact detection, we replace them with the average value of pixels around them, after continuum subtraction. Generally this works well, since it makes the bad pixel look exactly like the surrounding area, but occasionally this creates odd effects in the image too.
(image credit: Euclid)
What are the spiky Xs?
These are stars with diffraction spikes. Even though these are astronomical objects, they aren't the galaxy spectra the data was intended to observe, and the spikes are caused by the telescope. You might also find SAM struggles to mask these correctly!
When light comes into the telescope, it hits mirrors as it is reflected and focused at the camera. However, these mirrors have sharp edges, and machinery to support the mirrors, called struts. Light waves bend around these edges and struts, creating diffraction spikes. Each unique telescopes has a different diffraction spike pattern depending on the shape of the mirror and number of struts. See the image below for some examples.
(image credit: NASA)
What are the long, skinny white streaks?
Some of the images have long streaks across the frame. These artifacts may be due to cosmic rays. These are high energy particles flying through space that hit the detector. Some come from the solar wind off our own sun, but some come from further out in our galaxy! The length of the stripe in the image depends on the angle that the particle hits the telescope.
(image credit: Euclid)
What are the figure eight shaped blobs?
These are called zeroth order sources. They are extra images of a galaxy created by the instrument on the telescope.
What is a segmentation map?
A segmentation map is a way to "cut out" different objects in an image. The computer identifies different objects, and color codes them.
How do I tell if the SAM identified the artifact correctly?
The SAM identified the artifact correctly if the entire object highlighted by the green box, but no extra pixels is blue. However, this can be somewhat subjective if the edge is unclear or it is a dim object that doesn't stand out well from the background. Make your best estimate at which pixels should or shouldn't be highlighted.
How many pixels is too many extra pixels?
Any extra scattered pixels or pixels that are making a tail off of the main object are incorrect masks. If it is an extra layer of blue pixels surrounding the object (maintaining the artifact's overall shape) that's okay! It's better to have too many pixels in blue than too few. Mostly we don't want to mask double or triple the size of the artifact because then we'd be at risk of losing data we want to analyze later.
What telescope is this data from?
This ESA Euclid / Euclid Consortium /NASA Q1-2025 data is from the European Space Agency's Euclid mission.
When is Roman launching?
NASA's Nancy Grace Roman Space Telescope launched on August 30, 2026.
What is a machine learning model?
A machine learning model is code that is trained to find patterns in data and then use that information to make informed decisions on new data. Common uses of this are to classify images, or predicting an outcome based on input data. All AI in this project is just a machine learning model designed to work with images.
What is the AI used in this project?
The model we use is called the SAM, from Meta AI. It was trained on licensed and privacy protecting images. The results from your aggregated classifications (no single user) will be used by the Zooniverse researchers to improve this model for astronomical purposes only. We want to create better segmentation maps in the future for Roman.
Are you working with Meta? Will they see my data?
No. We are not associated in any way with Meta; we are just employing a model they developed. Any retraining of the model will be done independently from Meta and by Zooniverse researchers for astronomical purposes only. No volunteer identifiable information will be used, only aggregate classifications.