Automated Microfossil Identification and Segmentation using a Deep Learning Approach

Abstract Computational analysis applicability to paleontological images ranges from the study of the evolution of animals, plants and microorganisms to the habitat simulation of living beings from a specific epoch. It can also be applied in several niches, e.g. oil exploration, where several factors can be analyzed in order to minimize costs related to oil extraction. One specific factor is the characterization of the environment to be explored. This analysis can occur in several ways: use of probes, samples extraction, correlation with logs of other drilling wells and so on. During the samples extraction phase, the Computed Tomography (CT) is of extreme importance, since it preserves the sample and makes it available for several analyses. Based on 3D images generated by CT, analyses and simulations can be performed, and processes currently performed manually and exhaustively, can be automated. In this work, we propose and validate a method for fully automated microfossil identification and segmentation. A pipeline is proposed that begins with scanning and ends with the microfossil segmentation process. For the microfossil segmentation, a Deep Learning approach was developed, which resulted in a high rate of correct microfossil segmentation (98% IOU). The validation was performed both through an automated quantitative analysis and visual inspection. The study was performed on a limited dataset, but the results provide evidence that our approach has potential to be generalized to other carbonatic rock substrates. To the extent of the authors' knowledge, this paper presents the first fully annotated MicroCT acquired microfossils dataset made publicly available.