Upload PDF for Form Field Detection

Using vision, NOT OCR, I am passing an image of a form with red bounding boxes and a FIELD ID for each field found in the form.  ONLY for each FIELD ID WITH red bounding box, I need you to label with the appropriate label or question needed to fill it out.  If there is a FIELD ID with NO red bounding box, ignore FIELD ID.

IMPORTANT: There cannot be two labels with the same name. You do not need to extract x,y coordinates; instead, use the red bounding boxes to determine what label or question each field is associated with and the type of field it is.

For each red bounding box, provide the following information:

field: The printed FIELD ID for each field (e.g., Field 1, Field 2, etc.).
label: The label or question of the field. There cannot be two labels with the same name.
type:  The type of field it is (e.g., text, checkbox, date, etc) that will be used to create the appropiate acroform field later
value:  The printed FIELD ID for each field (e.g., Field 1, Field 2, etc.).

Start with "Field 1" and proceed sequentially, ensuring no field name is repeated. Output the fields in order, starting from the top left corner and moving to the bottom right corner.

Here is an example of the JSON response format/schema:

{
  "fields": [
    {
      "field": "Field 1",
      "label": "Razón o denominación social",
      "type": "checkbox",
    }
  ]
}

After extracting labels, check that 1) no two labels have the same values, and 2) all red bounding boxes have a corresponding label associated with it

Select ChatGPT Model:

Upload PDF for Form Field Detection

Upload supporting documents and vlm_label_results JSON

Upload acroform and merged_result JSON