Artificial intelligence applied to the classification of eight middle Eocene species of the genus <i>Podocyrtis</i> (polycystine radiolaria)

Carlsson, Veronica; Danelian, Taniel; Boulet, Pierre; Devienne, Philippe; Laforge, Aurelien; Renaudie, Johan

doi:https://doi.org/10.5194/jm-41-165-2022

Articles | Volume 41, issue 2

https://doi.org/10.5194/jm-41-165-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/jm-41-165-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 41, issue 2

Research article

|

04 Nov 2022

Research article |

| 04 Nov 2022

Artificial intelligence applied to the classification of eight middle Eocene species of the genus Podocyrtis (polycystine radiolaria)

Veronica Carlsson, Taniel Danelian, Pierre Boulet, Philippe Devienne, Aurelien Laforge, and Johan Renaudie

Supplement

https://doi.org/10.5194/jm-41-165-2022-supplement

Data sets

Podocyrtis Image Dataset Veronica Carlsson https://doi.org/10.57745/G7CHQL

Model code and software

Codes for image preparation and MobileNet CNN Veronica Carlsson and Aurelien Laforge https://doi.org/10.57745/J4YL4I

Short summary

This study evaluates the use of automatic classification using AI on eight closely related radiolarian species of the genus Podocyrtis based on MobileNet CNN. Species belonging to Podocyrtis are useful for middle Eocene biostratigraphy. Numerous images of Podocyrtis species from the tropical Atlantic Ocean were used to train and validate the CNN. An overall accuracy of about 91 % was obtained. Additional Podocyrtis specimens from other ocean realms were used to test the predictive model.