deepdoctection/deepdoctection · Can't get the layout similar to the app on Hugginface

Hi, I am new to document OCR and have been trying to create a Python script on my Linux machines for the last four weeks, but I can't get the layout results that the app on Hugginface (https://huggingface.co/spaces/deepdoctection/deepdoctection) is getting. For me, many text pieces are NOT covered under layout selection. I tried to modify several config parameters blindly but to no avail.

Also, the layout logic per page often identifies a huge "figure" instead of small segments. Please find the sample image below.

Huggingface Result:

MY Own Result:

So, my question is, HOW CAN I REPLICATE THE EXACT SETUP CREATED BY THE HUGGINGFACE APP in my local environment?
Please, please, please help!!!