Deep learning twined spatial analysis for detection of mysterious fairy circles

← Back to Research

19 November 2025 · First online (indexed publication record)
Hridoy Jyoti Mahanta, Virendra M Tiwari
Scientific reports · Computational image-screening study

Researchers trained image-classification models to locate circular landscape patterns in satellite imagery. The study presents a way to screen large areas for candidates requiring further investigation.

Method and scope

Pre-trained convolutional neural networks were applied to geomorphological imagery and evaluated for pattern detection.

What to keep in mind

A detected circular feature is not proof of natural hydrogen or a particular ecological mechanism. Different kinds of circular features require field verification.

Original paper & source

Read the original source record →
DOI: 10.1038/s41598-025-03691-4

Indexed original abstract, citation and linked publication-status metadata reviewed through Europe PMC; full text not assessed. UKPX reference added and checked 7 November 2026. The date above belongs to the paper.