storytracer/latin-ocr-sample

该数据集名为latin-ocr-sample，是一个用于OCR模型比较的数据集。每行数据将一个源页面与一个模型的标准化输出配对。数据集包含5个页面、4个模型，共20行数据。所有页面均无错误，且源图像已嵌入数据集中。数据集提供了每个模型的性能指标、页面间的差异度以及数据模式的详细说明。此外，还提供了如何使用该数据集的示例代码和浏览工具。

The dataset named latin-ocr-sample is designed for OCR model comparison. Each row pairs one source page with one models normalized output. It includes 5 pages, 4 models, and 20 rows in total. All pages are error-free, and source images are embedded in the dataset. The dataset provides detailed performance metrics for each model, disagreement scores between pages, and a comprehensive schema description. Additionally, it includes example code for loading the dataset and tools for browsing.