AI Spots 6,000 New Craters on the Moon


Moon surface by telescope / Photo by Procy via Shutterstock


Researchers from the University of Toronto Scarborough in Canada used artificial intelligence to automate the whole process of identifying craters on the moon. The technique invented by the researchers enabled them to spot 6,000 new moon craters that were previously unidentified in traditional manual counting.

Mohamad Ali-Dib, a postdoctoral fellow at U of T Scarborough’s Centre for Planetary Science, shared that without their new technique, astronomers would need to look manually at the image, spot, and then count the craters. Then they will calculate how large these craters are. With AI though, the entire process helps them saves a great amount of time and effort.

Ali-Dib developed the technique with Chenchong Charles Zhu, Ari Silburt, and a group of researchers from the Canadian Institute for Theoretical Astrophysics and CPS. The group first trained the neural network on a big data set that covers two-thirds of the moon. Then, they tested the trained network on the remaining portion of the moon. “It worked so well that it was able to identify twice as many craters as traditional manual counting,” stated the U of T Scarborough’s news.

Silburt, U of T’s former grad student in the Department of Astronomy and Astrophysics, said that the new craters spotted by the AI are unrealistic for humans to characterize efficiently by eye and that machines have real potential in revealing the “undiscovered clues” about the solar system’s formation.