This is an official implementation for the WTW Dataset in “Parsing Table Structures in the Wild ” on ICCV 2021. Here, you can download the paper, and Supplementary materials.

WTW-Dataset is the first wild table dataset for table detection and table structure recongnition tasks, which is constructed from photoing, scanning and web pages, covers 7 challenging cases like: (1)Inclined tables, (2) Curved tables, (3) Occluded tables or blurredtables (4) Extreme aspect ratio tables (5) Overlaid tables, (6) Multi-color tables and (7) Irregular tables in table structure recognition.


It contains 14581 images with the following ground-truths:

- data
 - train
  - images
  - xml (including image name, table id, table cell bbox(four vertices), start col/row, end col/row)
 - test
  - images
  - xml
  - class (7 .txt files include image names for 7 different challenging cases)

Download link is here

To be updated

Our results on WTW-dataset

Evaluation code

Data to other forms:

If you want to change to other common forms, you can do followings :

  • run the xmltococo.py to change the xml to json form.(To be updated)
  • run the xmltohtml.py to change the xml to html form.(To be updated)

Model link

Our model Cycle-Centernet has been used as Alibaba’s online business software, so we can’t open the model code. If you need to test, you can use the following online test link to try the different table images.


If you use the dataset, please consider citing our work-

	author = {Rujiao, Long and Wen, Wang and Nan, Xue and Feiyu, Gao and Zhibo, Yang and Yongpan, Wang and Gui-Song, Xia},
	title = {Parsing Table Structures in the Wild},
	booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
	month = {October},
	year = {2021}