Résumé
The main goal of the third year of the SnakeCLEF challenge was to provide an evaluation platform that helps track the performance of AI-driven methods for snake species recognition systems on a global scale and allows direct comparison with human experts. We ran two challenges separately for humans — experts and novices — and AI methods in order to lay the groundwork for future comparison between human and machine-based snake species identification. We have provided 187,129 snake observations with 318,532 photographs — 270,251 for training and 48,281 for testing — of 1,572 snake species collected in 208 countries. The human performance evaluation was conducted on a tailored subset with 150 images derived from the full test set. We report (i) a description of the provided data, (ii) evaluation methodology and principles, (iii) an overview of the methods submitted by the participating teams, and (iv) a discussion of the obtained results. © 2022 Copyright for this paper by its authors.