Public Domain Image Dataset for Ivory/Bone Object Classification

The dataset includes a selection of public domain images used for training a machine learning image segmentation algorithm to detect and classify ivory and bone objects. The images were used as a training and test datasets described in the following publication: Simone Mozzon, Andrius Tamosiunas, David Bacon, Caroline Cox, Alan Collins, Automated identification of ivory carvings using machine learning: A proof-of-concept approach to detecting the illegal wildlife trade, 2026, submitted for publication to PLOS One. We provide both the training and test dataset images along with the binary masks used in the training of the segmentation model. The following data files are available: bone_and_ivory.zip: bone and ivory images used for the training and validation datasets binary_masks.zip: binary masks used for each image in the training and validation dataset ivory.zip: the public domain ivory images used for testing the trained model bone.zip: the public domain bone images used for testing the traind model The dataset includes a wide selection of objects, ranging in type from ivory/bone figurines to museum pieces and decor objects. The images have different resolutions, image quality and lightning conditions. The images also have varying formats (primarily .jpg and .png). The images come from two sources: The private collection curated by Caroline Cox (University of Portsmouth) The Wikimedia Commons public domain image catalogue.