SCI-3000: A Novel Dataset for the Task of Figure, Table and Caption Extraction from Scientific PDFs

This dataset contains bounding boxes of figures, tables, captions in 34,791 pages extracted from 3000 open-access scientific publications from the fields of medicine, chemistry, physics, computer science, and technology. The underlying publications are also included in PDF form. For more details, refer to the README file.本数据集包含从医学、化学、物理、计算机科学与技术领域的3000篇开放获取科学出版物中提取的34791页内的插图、表格及对应题注的边界框（bounding box）。上述原始出版物亦以PDF格式一并提供。如需了解更多细节，请参阅README文件。