Mismatching of table headers

[vikas-singh16]

hi, thanks for making such a good repo for such a challenging problem.

I am encountering a constant issue of mismatching of table columns header. while exploring this issue in depth, I can across a fucntion called "build_table_from_html_and_cell".

I don`t know whether it is due to this function or something else.

Please can u tell me how is it you are merging the html output and cell output. What logic you are using?

For your reference
Input Image

Output

There are some cases where the result is much worse than this.

[yumikim381]

If i understood correctly there are as many cell output (text) as bounding boxes. So given a list of cell outputs and bounding boxes, the function just goes through the lists and pop one element each from both list. Maybe reading their paper helps to understand better.

[vikas-singh16]

Thanxs for the reply. but what i mean to say is if you look at the input image header and output image header, you will notice a mismatch of headers. Is there a way to fix it?

[yumikim381]

I m testing out this repo for next days as well, if i find anything that could help resolving this issue, i ll let you know :)

[gjgjh]

Although the number of cell output text and bbox are the same, the number of cells in the html structure output does not necessarily match the other two, which may result in the bad result.