Dataset Hub

PubLayNet is a large document image dataset whose layout annotations consist of bounding boxes and polygon segmentations. The data originate from the PubMed Central Open Access Subset, and annotations are generated by automatically matching PDF and XML formats. This dataset is the largest in the field of document layout analysis; see the paper "PubLayNet: largest dataset ever for document layout analysis" for details.

pdf-layout-chinese is a Chinese document layout analysis dataset focusing on Chinese scholarly documents (e.g., papers). The dataset provides 10 layout classes: text, title, image, image title, table, table title, header, footer, caption, and formula. It contains 5,000 training images and 1,000 validation images; each image has a correspondingly named JSON annotation file. Annotations were created with labelme and support polygon shapes.