PubLayNet: largest dataset ever for document layout analysis

TL;DR

PubLayNet is a large-scale dataset of over 360,000 document images with annotated layout elements, enabling more effective training of neural networks for document layout analysis, especially in scientific articles.

Contribution

The paper introduces PubLayNet, the largest publicly available dataset for document layout analysis, facilitating improved deep learning models for scientific document understanding.

Findings

Deep neural networks trained on PubLayNet accurately recognize scientific article layouts.

Pre-trained models on PubLayNet are effective for transfer learning in different document domains.

The dataset enables development of more advanced document layout analysis models.

Abstract

Recognizing the layout of unstructured digital documents is an important step when parsing the documents into structured machine-readable format for downstream applications. Deep neural networks that are developed for computer vision have been proven to be an effective method to analyze layout of document images. However, document layout datasets that are currently publicly available are several magnitudes smaller than established computing vision datasets. Models have to be trained by transfer learning from a base model that is pre-trained on a traditional computer vision dataset. In this paper, we develop the PubLayNet dataset for document layout analysis by automatically matching the XML representations and the content of over 1 million PDF articles that are publicly available on PubMed Central. The size of the dataset is comparable to established computer vision datasets, containing over 360 thousand document images, where typical document layout elements are annotated. The experiments demonstrate that deep neural networks trained on PubLayNet accurately recognize the layout of scientific articles. The pre-trained models are also a more effective base mode for transfer learning on a different document domain. We release the dataset (https://github.com/ibm-aur-nlp/PubLayNet) to support development and evaluation of more advanced models for document layout analysis.