DOCTRACK
收藏资源简介:
DOCTRACK是由上海大学计算机工程与科学学院创建的一个视觉丰富文档(VRD)数据集,旨在通过眼动追踪技术模拟人类阅读顺序。该数据集包含409个样本,涵盖多种文档类型,如表格、图表和图形,用于研究机器阅读VRD的挑战。数据集的创建过程涉及从现有数据集中随机选择文档,并通过眼动追踪实验收集人类阅读顺序数据。DOCTRACK的应用领域主要集中在改进文档AI模型,以更准确、连续和灵活地阅读VRD,解决现有模型在处理复杂文档结构时的局限性。
DOCTRACK is a visually rich document (VRD) dataset created by the School of Computer Engineering and Science, Shanghai University. It aims to simulate human reading sequences via eye-tracking technology. This dataset contains 409 samples covering various document types such as tables, charts and graphics, and is used to study the challenges of machine reading of VRDs. The dataset creation process involves randomly selecting documents from existing datasets and collecting human reading sequence data through eye-tracking experiments. The application scenarios of DOCTRACK mainly focus on improving document AI models to read VRDs more accurately, continuously and flexibly, so as to address the limitations of existing models when dealing with complex document structures.




