Brno Mobile OCR Dataset (B-MOD)

Brno Mobile OCR Dataset (B-MOD) 是由布尔诺理工大学信息学院创建的一个专注于移动设备拍摄的低质量文档光学字符识别（OCR）数据集。该数据集包含2113个来自科学论文的独特页面，通过23种不同的移动设备拍摄，总计19725张照片，每张照片都附有精确的位置和50万条文本行注释。B-MOD数据集的创建过程涉及使用增强现实（AR）标记进行页面定位和精确的文本行注释。该数据集主要用于开发和评估针对低质量图像的文档分析方法，特别适用于文本行级别的识别、定位、布局分析、图像修复和文本二值化等领域。

Brno Mobile OCR Dataset (B-MOD) is an optical character recognition (OCR) dataset focused on low-quality document images captured by mobile devices, developed by the Faculty of Information Technology at Brno University of Technology. It contains 2,113 unique pages extracted from scientific papers, captured with 23 different mobile devices, totaling 19,725 photographs. Each photograph is paired with precise positional annotations and 500,000 text line annotations. The creation of the B-MOD dataset involved using augmented reality (AR) markers for page positioning and accurate text line annotation. This dataset is primarily used for developing and evaluating document analysis methods for low-quality images, and is particularly applicable to tasks such as text line-level recognition, localization, layout analysis, image inpainting, and text binarization.