Deformation Wired Table (DWTAL)
收藏资源简介:
DWTAL数据集是为了解决变形表格的单元格定位问题而创建的。该数据集由南昌航空大学的研究团队开发,包含了从TAL-OCR和WTW数据集中选取的轻微变形表格,经过数据生成器处理,生成了大量变形表格数据。数据集被分为两个独立的部分:DWTAL-s和DWTAL-l,分别包含8765张和19520张变形表格。这些数据集为变形表格的单元格定位提供了丰富且精细的空间坐标信息,支持像素级分割模型的训练,有助于解决下游任务中的内容获取问题。
The DWTAL dataset was developed to tackle the issue of cell localization for deformed tables. Developed by the research team at Nanchang Hangkong University, this dataset includes slightly deformed tables selected from the TAL-OCR and WTW datasets, and a substantial amount of deformed table data was generated through a dedicated data generator. The dataset is divided into two separate subsets: DWTAL-s and DWTAL-l, which respectively contain 8765 and 19520 deformed table images. These datasets provide rich and precise spatial coordinate information for cell localization of deformed tables, enabling the training of pixel-level segmentation models and facilitating the resolution of content acquisition issues in downstream tasks.
OG-HFYOLO数据集概述
数据集基本信息
- 数据集名称:OG-HFYOLO
- 主要用途:变形表格单元格实例分割
- 相关论文:arXiv:2504.20682
数据集格式与存储
- 主要存储格式:YOLO格式
其他格式支持
数据转换工具
- 转换脚本位置:./dataprocess
- 包含脚本:
- yolo2coco.py
- yolo2labelme.py
- labelme2yolo.py
- labelme2coco.py
- 包含脚本:
- 数据生成器:./dataprocess/data_gen.py
使用要求
-
Python版本:≥3.8.0
-
PyTorch版本:≥1.8
-
依赖安装:
pip install -r requirements.txt 或 conda env create -f environment.yaml
许可信息
- 许可证类型:GNU Affero General Public License v3.0 (AGPL-3.0)

- 1OG-HFYOLO :Orientation gradient guidance and heterogeneous feature fusion for deformation table cell instance segmentation南昌航空大学 · 2025年



