Accuracy of a neural net classification of closely-related species of microfossils from a sparse dataset of unedited images

Abstract

Identification of biologic objects in images is a major source of biodiversity data. Currently this is done by scarce taxonomic experts and data is thus limited in scope and reproducibility. Automated identification in fields such as plankton research or micropaleontology, where enormous numbers of objects are available, would significantly improve data quantity and quality, particularly in applied studies of environmental and climate change. We describe a machine learning workflow based on the MobileNet convolutional network. The software can identify closely related species of radiolarians, a morphologically challenging group of microfossils, and from complete species populations (not only ideal specimens) as they are normally identified in standard transmitted light microscope preparations. Multiple, partial focus, depth of field limited images were obtained for each fossil specimen from multiple radiolarian microslides. Images were normalized and in one test also cropped to remove most systematic slide-linked image biases (e. g. type of background particles) that could be used by a classifier as non-taxonomic clues to species assignment. An average of 60 specimens per species for 16 species in two distinct clusters of closely related forms (9 species in the Antarctissa group and 7 species in the genus Cycladophora) were used to train and test the system. An overall average classification accuracy of ca 73% was achieved, and for some species >85%. Using a cutoff for specimens with classifier-calculated low certainty values boosts overall accuracy close to 90%, but at the cost of ca 1/3 reduction in identifiable specimens. This latter accuracy is close to the reproducibility of human experts, albeit with more unidentifiable specimens. The most important constraint to broader use is the time and effort needed by taxonomic experts to collect and label images to be used in training, as many species in these diverse biotas are rare, and the numbers of taxonomic experts available are very limited.

Automatic taxonomic identification based on the Fossil Image Dataset (>415,000 images) and deep convolutional neural networks

“…The taxonomic positions of some taxa may be modified in subsequent studies, it being necessary to coordinate classification criteria among scholars and research communities (Al-Sabouni et al 2018; Fenton et al 2018) for consistency. Even internal variations of taxa in different regions and periods should be considered, and DCNNs can also be used to verify these problems (Pires de Lima et al 2020), given their highly accurate, reproducible, and unbiased classification (Renaudie et al 2018; Hsiang et al 2019; Marchant et al 2020).…”

Technical note: A new automated radiolarian image acquisition, stacking, processing, segmentation and identification workflow

“…However, despite their usefulness for such investigations, radiolarians are not as utilized in the same way as other microfossil groups, such as benthic and planktic foraminifera, or nannofossils, such as coccolithophorids. Experts on living and fossil radiolarians are relatively scarce, and some radiolarian species still lack a satisfactory taxonomy, especially for taxa within the order Spumellaria (Riedel, 1967;Sanfilippo et al, 1985). Identification of a substantial and sufficient number of specimens per sample (usually about 300 for reliable assemblage composition estimations; Fatela and Taborda, 2002) is very time-consuming and requires a consistent and detailed taxonomic knowledge.…”

“…In 2017, Keceli et al (2017) investigated scanning electron microscope (SEM) images of 27 selected Triassic species. Renaudie et al (2018) recently achieved promising results focusing on the automated identification of species from the same genus with transmitted light microscope images. They obtain an overall identification accuracy of 73 %, achieved over 16 species from 2 genera, where the morphological difference between species can be very tricky.…”