| Popis: |
Multi- elemental and -dimensional data are more and more important during the development of data-driven research, as is the case in modern palaeontology, in which visual examination, by experts or someday the artificial intelligence, to every fossil specimen acts a crucial and fundamental role. We here release an integrated image dataset of 113 Ordovician to Silurian graptolite species or subspecies that are significant in global stratigraphy and shale gas exploration. The dataset contains 1550 high-resolution graptolite specimen images and scientific information related to the specimen, e.g., every specimen's taxonomic, geologic, geographic, and related references. We develop a tool, FSIDvis (Fossil Specimen Image Dataset Visualiser), to facilitate the human-interactive exploration of the rich-attribution image dataset. A nonlinear dimension reduction technique, t-SNE (t-Distributed Stochastic Neighbor Embedding), is employed to project the images into the two-dimensional space to visualise and explore the similarities. Our dataset potentially contributes to the analysis of the global biostratigraphic correlations and improves the shale gas exploration efficiency by developing an image-based automated classification model. All images are available from https://doi.org/10.5281/zenodo.5205216 (Xu, 2021). |
