An image dataset of fusulinid foraminifera generated with the aid of deep learning

Huang, Hongyi; Shi, Yukun; Chen, Qin; Xu, Hongbo; Song, Sicong; Shi, Yujie; Shen, Furao; Fan, Junxuan

doi:10.1002/gdj3.215

Cited by 3 publications

(7 citation statements)

References 67 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Comparing to the main experiments over the larger dataset (Huang et al, 2023), more considerable improvements of OGS against OOO are presented for the experiments over this smaller dataset (Pires de Lima et al, 2020) when taking Inception-V4 (+3% for ACC@1 and +2.84% for Macro-F1) and RegnetY (+3.78% for ACC@1 and +4.51% for Macro-F1) as the base in Table 3, though OGS has shown considerable improvements when taking ResNet-50 (+1.42%…”

Section: High Applicability To Small Datasetsmentioning

confidence: 99%

“…The main dataset used in this work is 'Fusulinid images 2400 -NJU', which is described in detail in Huang et al (2023) The 2400 images are selected evenly from 16 genera of all six fusulinid families: Fusulinidae, Schwagerinidae, Ozawainellidae, Schubertellidae, Neoschwagerinidae and Verbeekinidae (see Table 1). Images of holotypes, paratypes, cotypes and syntypes of the selected species are preferably chosen as they better represent the described morphological features.…”

Section: Datasetmentioning

confidence: 99%

“…Although the images are labelled to the species level, using this level would result in a significantly imbalanced data volume, so the genus level was chosen in our study. In the main experiments, the Huang et al (2023) dataset is split into a training set, a validation set and a test set, with 110, 20 and 20 images for each genus as the default setting.…”

Section: Datasetmentioning

confidence: 99%

“…Although the colour may somewhat reflect differences in the living and preservation environments of various classes of fossils, the optical microscope used, the filming equipment and parameters, and the factors of printing and scanning may also introduce colour-related noise. For the main dataset Huang et al (2023), there is little difference in whether the colour is included or not, as the performance of O and G does not show much comparable variance (see Table 2). However, for the Pires de Lima et al (2020) dataset as shown in 2.…”

Section: High Applicability To Small Datasetsmentioning

confidence: 99%

“…To investigate the effectiveness of the newly proposed method in this study, we utilize the largest dataset of fusulinids to date (Huang et al, 2023), containing 2400 images from 16 genera that cover all six fusulinid subfamilies with respect to the classification system of Sheng et al (1988).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Fossil image identification using deep learning ensembles of data augmented multiviews

Hou,

Lin,

Huang

et al. 2023

Methods Ecol Evol

Self Cite

View full text Add to dashboard Cite

Identification of fossil species is crucial to evolutionary studies. Recent advances from deep learning have shown promising prospects in fossil image identification. However, the quantity and quality of labelled fossil images are often limited due to fossil preservation, conditioned sampling and expensive and inconsistent label annotation by domain experts, which pose great challenges to training deep learning‐based image classification models. To address these challenges, we follow the idea of the wisdom of crowds and propose a multiview ensemble framework, which collects Original (O), Grey (G) and Skeleton (S) views of each fossil image reflecting its different characteristics to train multiple base models, and then makes the final decision via soft voting. Experiments on the largest fusulinid dataset with 2400 images show that the proposed OGS consistently outperforms baselines (using a single model for each view), and obtains superior or comparable performance compared to OOO (using three base models for three the same Original views). Besides, as the training data decreases, the proposed framework achieves more gains. While considering the identification consistency estimation with respect to human experts, OGS receives the highest agreement with the original labels of dataset and with the re‐identifications of two human experts. The validation performance provides a quantitative estimation of consistency across different experts and genera. We conclude that the proposed framework can present state‐of‐the‐art performance in the fusulinid fossil identification case study. This framework is designed for general fossil identification and it is expected to see applications to other fossil datasets in future work. Notably, the result, which shows more performance gains as train set size decreases or over a smaller imbalance fossil dataset, suggests the potential application to identify rare fossil images. The proposed framework also demonstrates its potential for assessing and resolving inconsistencies in fossil identification.

show abstract

Section: High Applicability To Small Datasetsmentioning

confidence: 99%

Section: Datasetmentioning

confidence: 99%

Section: Datasetmentioning

confidence: 99%

Section: High Applicability To Small Datasetsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Fossil image identification using deep learning ensembles of data augmented multiviews

Hou,

Lin,

Huang

et al. 2023

Methods Ecol Evol

Self Cite

View full text Add to dashboard Cite

show abstract

Artificial intelligence in paleontology

Yu,

Qin,

Watanabe

et al. 2024

Earth-Science Reviews

View full text Add to dashboard Cite

AI in paleontology

Qin

Watanabe³

et al. 2023

Preprint

View full text Add to dashboard Cite

Accumulating data have led to the emergence of data-driven paleontological studies, which reveal an unprecedented picture of evolutionary history. However, the fast-growing quantity and complication of data modalities make data processing laborious and inconsistent, while also lacking clear benchmarks to evaluate data collection and generation, and the performances of different methods on similar tasks. Recently, Artificial Intelligence (AI) is widely practiced across scientific disciplines, but has not become mainstream in paleontology where manual workflows are still typical. In this study, we review more than 70 paleontological AI studies since the 1980s, covering major tasks including micro- and macrofossil classification, image segmentation, and prediction. These studies feature a wide range of techniques such as Knowledge Based Systems (KBS), neural networks, transfer learning, and many other machine learning methods to automate a variety of paleontological research workflows. Here, we discuss their methods, datasets, and performance and compare them with more conventional AI studies. We attribute the recent increase in paleontological AI studies to the lowering bar in training and deployment of AI models rather than real progress. We also present recently developed AI implementations such as diffusion model content generation and Large Language Models (LLMs) to speculate how these approaches may interface with paleontological research. Even though AI has not yet flourished in paleontological research, successful implementation of AI is growing and show promise for transformative effect on the workflow in paleontological research in the years to come.

show abstract

An image dataset of fusulinid foraminifera generated with the aid of deep learning

Cited by 3 publications

References 67 publications

Fossil image identification using deep learning ensembles of data augmented multiviews

Fossil image identification using deep learning ensembles of data augmented multiviews

Artificial intelligence in paleontology

AI in paleontology

Contact Info

Product

Resources

About