GROTOAP2

GROTOAP2 (GROund Truth for Open Access Publications) is a dataset useful for training and performance evaluation of document content analysis tasks, such as document zone classification. GROTOAP2 is a successor of GROTOAP dataset. GROTOAP2 was built automatically from PubMed Central Open Access Subset. It contains 13,210 ground truth files, that store geometrical and logical structure of the articles content. The corresponding PDF files can be downloaded from PMC repository using provided script. This repository contains a sample of a 132 ground truth files. The full dataset can be downloaded from: http://cermine.ceon.pl/grotoap2/.