Researchers use deep transfer learning to study nest site fidelity in painted stork in Delhi zoo

A team of researchers have successfully used deep transfer learning (DTL), a non-invasive artificial intelligence-based tool, to study nest site fidelity in painted stork (Mycteria leucocephala) in the National Zoological Park, Delhi.

The researchers chose a male painted stork with a distinctive injury scar on its neck and observed it over four consecutive breeding seasons from 2022 to 2025 to assess the nest site fidelity – a trait of using the same nesting site in successive breeding seasons.

They named the stork ‘Ringo’, in honour of drummer Ringo Starr of iconic pop music band The Beatles, for the study. A total of 2,349 high resolution images of Ringo, covering both sides and wing markings in the folded position, were photographed for the study. They also clicked 1,755 images, showing both the left and right sides of the wings of other nesting storks. These images collected during breeding seasons from 2022 to 2025 were used as a foundation for studying individual identification through unique features present in painted storks.

As per the study, published in Royal Society Open Science, the researchers employed a non-invasive approach to monitor Ringo. They used scale-invariant feature transform (SIFT), a computer vision algorithm that extracts distinctive features from images for accurate matching and individual identification. The SIFT features identified the unique scar marking in Ringo images.

The researchers also developed a DTL model, to identify features that distinguish Ringo from other storks, with the feather pattern serving as a form of biological fingerprint. The tool validated Ringo’s identity with 98% accuracy and the bird’s repeated sightings at the same spot over four consecutive years confirmed its nest-site fidelity.

The study was conducted by a team comprising Abdul Jamil Urfi and Paritosh Ahmed from the Department of Environmental Studies, University of Delhi; Mylswamy Mahendiran from the Division of Wetland Ecology, Salim Ali Centre for Ornithology and Natural History, Coimbatore, and Mylswamy Parthiban from Department of Physical Sciences and Information Technology, Agricultural Engineering College and Research Institute, Tamil Nadu Agricultural University, Coimbatore.

As per the study, the findings highlight the potential of pattern-based recognition and use of DTL as a powerful and non-intrusive tool for long-term monitoring of colonial waterbirds.