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

This is a Fossil Image Dataset, which contains >415000 images. A total of 50 clades were labeled, with a final 90% accuracy. We used the web crawler to download fossil images from the Internet. We declare that all the collected images are used for academic purposes only. If anyone wants to use this dataset, please agree on the Terms of access for the Fossil Image Dataset (FID). We uploaded two datasets: FID (contains 0.415 million images) and reduced-FID (60 thousand images, 1200 for each clade). Requirements of necessary preinstalled Python libraries, algorithms for analysis, and the model weights are available at https://github.com/XiaokangLiuCUG/Fossil_Image_Dataset.