LAM数据集

LAM数据集是由意大利摩德纳和雷焦艾米利亚大学创建的大型线级手写文本识别数据集，专注于意大利古代手稿。该数据集包含25,823行文本，来源于单一作者在60年间的书信。数据集设计了两种配置：基本分割和基于日期的分割，旨在研究手写风格随时间的变化，以及在训练数据不可用的时间段内识别同一作者的文本。数据集的标注由两位专家手动完成，并进行了双重检查，确保了高精度。LAM数据集不仅适用于手写文本识别研究，还适用于研究手写风格随时间的变化。

The LAM dataset is a large-scale line-level handwritten text recognition dataset developed by the University of Modena and Reggio Emilia, Italy, focusing on ancient Italian manuscripts. It comprises 25,823 text lines sourced from letters penned by a single author across a 60-year period. The dataset provides two experimental configurations: a basic split and a date-based split, designed to investigate temporal variations in handwriting styles, as well as the recognition of texts from the same author during time periods where no training data is available. All annotations of the LAM dataset were manually completed by two experts and double-checked, ensuring high annotation accuracy. The LAM dataset is applicable not only to handwritten text recognition research, but also to studies on temporal changes in handwriting styles.