Individually recognising birds through the pattern of their plumage is an arduous task for humans. But not for the machine! An international collaboration involving mainly scientists from CNRS, the University of Montpellier1and the University of Porto in Portugal, has just shown how computers can learn to differentiate individuals from the same species of birds. These results are published on July 27, 2020 in Methods in Ecology and Evolution.
Recognition of different individuals of the same species is essential for studying wildlife populations, their adaptation process and behaviour. For the first time ever, research teams led by scientists from the Centre for Functional and Evolutionary Ecology (CNRS/University of Montpellier/Paul-Valéry-Montpellier University/ IRD/EPHE) and CIBIO at the University of Porto2 have individually identified birds using artificial intelligence techniques.
Researchers have developed a technique that allows them to collect a large number of bird shots: identified individuals with electronic marks have been photographed from several angles. This collection fed the computers who then learned to recognize birds by analyzing the images by methods of deep learning. Such a device has allowed computers to distinguish individuals according to patterns on their plumage, a task that humans cannot perform. The machines were able to differentiate individuals in three species of birds, the social republican, the coal and the Mandarin Diamond.
This new process can not only lead to less invasive methods of identification of wild individuals, but also to new knowledge of ecology. AI could also open up new possibilities in the study of animal behaviour in natural populations.
Photograph of a coal that illustrates how individual identification is done by computers. © André Ferreira
Deep learning-based methods for individual recognition in small birds, André C. Ferreira, Liliana R. Silva, Francesco Renna, Hanja B. Brandl, Julien P. Renoult, Damien R. Farinec, Rita Covas and Claire Doutrelant, Methods in Ecology and Evolution on July 27, 2020. DOI:10.1111/2041-210X.13436
- Working at the Centre for Functional and Evolutionary Ecology (CNRS/ University of Montpellier/ Paul-Valéry-Montpellier University/ IRD/ EPHE)
- This work was carried out in association with three other laboratories: the Paris-Saclay Institute of Neuroscience (CNRS/University Paris-Saclay), the Max Planck, and the FitzPatrick Institute in South Africa.