Large-Scale Historical Watermark Recognition

本数据集名为'大规模历史水印识别'，由法国国家档案学院创建，包含超过6000张新照片，用于支持一击实例识别和跨域一击实例识别等任务。数据集通过访问公共公证档案收集，涵盖了从14世纪到17世纪的大量水印样本，每类水印有多个实例，用于训练和评估深度学习算法在细粒度和跨域水印识别上的性能。该数据集首次允许在大规模上解决由水印定义的细粒度识别问题，对历史文档的分析和评估具有重要实际意义。

This dataset, named *Large-Scale Historical Watermark Recognition*, was created by the École Nationale des Chartes. It contains over 6,000 newly collected photographs to support tasks such as one-shot instance recognition and cross-domain one-shot instance recognition. The dataset was compiled by accessing public notarial archives, covering a large number of watermark samples spanning the 14th through the 17th century, with multiple instances available for each watermark category. It is intended for training and evaluating the performance of deep learning algorithms on fine-grained and cross-domain watermark recognition tasks. This dataset enables, for the first time, large-scale solutions to the fine-grained recognition problem defined by watermarks, holding significant practical implications for the analysis and evaluation of historical documents.