Linetext positions in percentage?

OCR.space Free OCR API
1

Hi, I see that the JSON results are divided by linetext and you provide coordinates (top, left) and width and height ll in pixels.

However, since I don’t know the size, in pixels, of the document, it’s hard to calculate where I should get what text to saved in a structured way. I would think that providing top, left, width, and height as
percentages would be more appropriate.

How other users are handling that?

Or how can I get the height and width of the document, so I can make those calculations?

2

What documents do you send? PNG? JPG? PDF?

3

I sent PNG and JPG documents

4

Here the returned coordinates are exactly the x/y pixels in the PNG/JPG files.

size, in pixels, of the document

This is the size of the images that you send. This works well for me.

5

Can you explain how you parse the values to variables.

I created a template JSON file from one document and I created a transform (translate) rule in which I get the most at left Line and most Top Line, as well as the maxLeft value for a line and the maxTop value too.

Using those, I created the following interpolation:
const diffX = maxLeft - globalLeft
const diffY = maxTop - globalTop

and on a loop I calculate
const factorX = (docVars[key].left - globalLeftModel)/diffXModel
const calcLeft = diffX * factorX + globalLeft

      const factorY = (docVars[key].top - globalTopModel)/diffYModel
      const calcTop = diffY * factorY + globalTop

where docVars is a JSON template as

const docVars = {"restriccion_movilidad": {"left": 31 , "top": 45},
                    "blindaje": {"left": 415, "top": 45}, 
                    "potencia": {"left": 527, "top": 45},
                    "declaracion": {"left": 32, "top": 101},
                    "ie": {"left": 382, "top": 98, "elem": 0},
                    "fecha_import": {"left": 382, "top": 98, "elem": 1},
                    "puertas": {"left": 604, "top": 100},
                    "limitacion_a_la_propiedad": {"left": 32, "top": 180},
                    "fecha_matricula": {"left": 32, "top": 248},
                    "fecha_exp": {"left": 242, "top": 248},
                    "fecha_vencimiento": {"left": 480, "top": 248},
                    "organismo_transito": {"left": 31, "top": 321},
                    "id": {"left": 376, "top": 555},
                    "minLeft": 31,
                    "diffX": 596
                  }

So, docVars[key].left for the first element would be 31 for (key) restriccion_movilidad and so on.
diffXModelis the at most left value on the whole template, and diffXModel is the highest left value on the template - diffXModel

calcLeft and calcTop are the interpolated values for the image I am processing at a given times.
Then, I iterate to the Lines and see which values are close to those calcLeft and calcTop.

It works more or less, but it isn’t very precise.

Anyone has a better approach?