Add MarkItDown in the benchmark

[hongbo-miao]

it would be great to add the new https://github.com/microsoft/markitdown in the benchmark, thanks! ☺️

[ouyanglinke]

Hi, we tried to run the model infer for MarkItDown but got empty results. Please let us know if there are any issues in the infer code.

Here is the moder infer code:

from markitdown import MarkItDown
import os

img_folder = './OmniDocBench/images'
save_path = './result0106/markitdown'
md = MarkItDown()

for img_name in os.listdir(img_folder):
    result = md.convert(os.path.join(img_folder, img_name))
    response = result.text_content
    with open(os.path.join(save_path, img_name[:-4] + '.md'), 'w', encoding='utf-8') as output_file:
        output_file.write(response)

[hongbo-miao]

Hi @ouyanglinke I haven't got chance to try markitdown myself. It is quite new.