Assessing the conservation status and extinction risk of species is a major challenge for biodiversity management. Currently, this assessment is done species by species, a process that requires in-depth expertise and is time-consuming. The ultimate goal of the CACTUS exploratory action is to be able to automatically predict these conservation statuses by automatically analyzing the masses of data available using artificial intelligence approaches. Indeed, recent cyberinfrastructures and advances in deep learning offer unprecedented opportunities to jointly analyze complex and heterogeneous biological data sources such as species occurrences, climatic and environmental data or satellite images. We are particularly interested in the use of participatory science data (such as those of the Pl@ntNet platform) which have the advantage of offering a high frequency of observation. To achieve this goal, however, we will have to solve difficult problems such as the lack of absence data, observation biases and the strong imbalance of available data between different species.
Inria teams involved
In partnership with
Univ Montpellier, CIRAD, IUCN, Caltech, Cornell Tech
© Inria / Photo C. Morel