This is a image dataset for deep learning studies. The main dataset comprises labeled CT slices from 3 protoceratopsian dinosaur skulls discovered from the Gobi Desert, Mongolia. The fossil specimens are now in the Division of Paleontology, American Museum of Natural History, New York, NY, USA. Inside the folders named after specimen number (i.e. IGM100-1021), there are six sub-folders comprising labeled or unlabled CT slice images of the specimen from three directions: axial, coronal, and sagittal. Each direction should contain all the information to reconstruct the entire specimen.

The other folder named Gobi Vert comprise CT slices, but not enough to reconstruct the original specimens, of multiple vertebrate fossils also discovered from the Gobi Desert, Mongolia. The specimen number and taxa are noted in the image names.

The CT scans were done in either the Department of Earth & Planetary Sciences, Yale University, USA or the American Museum of Natural History. The segmentation was done by authors, then labelled images are exported as the groundtruth for deep learning training. The detail of the dataset is following:

However, users are free to divide training, testing, and validating dataset as they need. 

The dataset comprises three rarely preserved embryonic protoceratopsian dinosaurs skulls, thus the morphological information embedded are also important for paleontological studies. The slices included here are enough to reconstruct the entire morphology of these embryonic skulls, therefore, if anyone would like to study the morphology of these fossils, please ask the consent from the correspondence author Congyu Yu cyu@amnh.org.