Qwen2.5-VL icon indicating copy to clipboard operation
Qwen2.5-VL copied to clipboard

Question - What is a good prompt to get the output of the image in a structured format.

Open AkshataABhat opened this issue 2 weeks ago • 2 comments

Since qwen2.5 VL is capable of maintaining layout information; I was wondering what is a good prompt to get output in a structured format. For example, given image of an invoice, I would like (specifically the 7B model) to output something like:

                                      Invoice ® ®
                     16 June 2025                                
                     Invoice No. 12345                        


  BILL TO
  Marceline Anderson
  +123-456-7890


DESCRIPTION                     PRICE          SUBTOTAL
---------------------------------------------------------
30x  Social Media Pack Design   $20.00        $600.00
5x   Furniture                  $100.00       $500.00
1x   Interior Design            $700.00       $700.00
1x   Architecture               $1000.00      $1000.00


 SUBTOTAL                       $2800.00  
 TAX                            $560.00  
 TOTAL                          $3360.00  

I have experimented with multiple prompts, but it seems like the model is unable to do this. Grateful for anyone who can guide me.

AkshataABhat avatar Feb 12 '25 04:02 AkshataABhat