DEV Community

Cover image for Rotation-invariant convolutional neural networks for galaxy morphologyprediction
Paperium
Paperium

Posted on • Originally published at paperium.net

Rotation-invariant convolutional neural networks for galaxy morphologyprediction

AI that reads galaxy shapes from any angle — faster than people

Imagine sorting millions of photos of far away star cities, but without looking at each one.
Astronomers usually rely on volunteers or experts to classify how a galaxy looks, and that takes a lot of time and it wont scale.
A new approach uses AI that learns to spot the shape no matter how the galaxy is turned, so it sees the same galaxy when rotated or moved.
This makes it great at finding spirals, disks and odd ones among billions of pixels.
For clear cases the program matches human choices with near 99% accuracy, which means experts only need check the tricky images.
That saves huge time on big sky surveys and frees people to study the interesting finds.
The tool handles many images fast, and its clever handling of rotation and movement helps it ignore useless differences.
Soon this will let scientists study millions of galaxies much faster, and hopefully reveal new surprises about how they form and change.

Read article comprehensive review in Paperium.net:
Rotation-invariant convolutional neural networks for galaxy morphologyprediction

🤖 This analysis and review was primarily generated and structured by an AI . The content is provided for informational and quick-review purposes.

Top comments (0)