Tibetan-header-footer
收藏资源简介:
Tibetan Header/Footer 是一个用于目标检测的数据集,专门标注了与页眉/页脚相关的页面布局元素。数据集包含313张页面图像,其中251张用于训练,62张用于验证。图像以YOLO文本标签格式标注,包含四个类别:页眉(header)、页脚(footer)、脚注(footnote)和页码(page_number)。数据集通过模型推理(使用DocLayout-YOLO和PP-DocLayout)生成初始边界框,随后通过Ultralytics Hub进行手动校正。每个图像文件(.jpg)都有对应的标签文件(.txt),且图像可能包含多个类别的边界框。数据集采用MIT许可证发布。
Tibetan Header/Footer is an object detection dataset specifically annotated for page layout elements related to headers and footers. The dataset comprises 313 page images, with 251 allocated for training and 62 for validation. All images are annotated in YOLO text label format, covering four categories: header, footer, footnote, and page_number. The initial bounding boxes of the dataset were generated through model inference using DocLayout-YOLO and PP-DocLayout, followed by manual correction via Ultralytics Hub. Each .jpg image file has a corresponding .txt label file, and images may contain bounding boxes of multiple categories. The dataset is released under the MIT License.



