Using conformal prediction for novelty detection in microfossil analysis

Presenter: Iver Martinsen, Doctoral Research Fellow, SFI Visual Intelligence/UiT – The Arctic University of Norway

Abstract: Microfossil analysis is crucial for subsurface mapping, such as matching strata between wells. Traditionally, specialist geoscientists manually examine numerous samples to identify key microfossil species. The Norwegian Offshore Directorate's digitalization of microfossil samples, combined with AI advancements, offers new possibilities for automating analysis. Unsupervised representation learning, a key AI research area, can generate useful image representations from large datasets without labels. While previous work has shown promising results for classifying limited classes, challenges remain in realistic settings with unknown species. Our recent work integrates unsupervised representation learning and uncertainty estimation to automate fossil analysis. We detail our approach and findings in three parts: training AI models using self-supervised learning methods, developing a conformal prediction method to manage diverse data, and creating distribution charts of fossils across multiple wells.

In compliance with GDPR consent requirements, presentations given in a Visual Intelligence context may be recorded with the consent of the speaker. All recordings are edited to remove all faces, names and voices of other participants. Questions and comments by the audience will hence be removed and will not appear in the recording.  With the freely given consent from the speaker, recorded presentation may be posted on the Visual Intelligence YouTube channel.

‍

‍

‍