I've been working on a project in which I extract table data from a pdf with neural network, I successfuly detect tables and get their coordinate (x,y,width,height) , I've been trying to crop the pdf with pypdf2 to isolate the table but for some reason cropping never matches the desired outcome. After running inference i get these coordinates
[[5.0948269e+01, 1.5970685e+02, 1.1579385e+03, 2.7092386e+02 9.9353129e-01]]
the 5th number is my neural network precision , we can safely ignore it
trying them in pyplot works , so there's no problem with them:
However using the same coords in pypdf2 is always off
from PyPDF2 import PdfFileWriter, PdfFileReader
with open("mypdf.pdf", "rb") as in_f:
input1 = PdfFileReader(in_f)
output = PdfFileWriter()
numPages = input1.getNumPages()
for i in range(numPages):
page = input1.getPage(i)
page.cropBox.upperLeft = (5.0948269e+01,1.5970685e+02)
page.cropBox.upperLeft = (1.1579385e+03, 2.7092386e+02)
output.addPage(page)
with open("out.pdf", "wb") as out_f:
output.write(out_f)
This is the output I get :
thank you !
Here you go:
from PyPDF2 import PdfFileWriter, PdfFileReader
with open("mypdf.pdf", "rb") as in_f:
input1 = PdfFileReader(in_f)
output = PdfFileWriter()
numPages = input1.getNumPages()
x, y, w, h = (5.0948269e+01, 1.5970685e+02, 1.1579385e+03, 2.7092386e+02)
page_x, page_y = input1.getPage(0).cropBox.getUpperLeft()
upperLeft = [page_x.as_numeric(), page_y.as_numeric()] # convert PyPDF2.FloatObjects into floats
new_upperLeft = (upperLeft[0] + x, upperLeft[1] - y)
new_lowerRight = (new_upperLeft[0] + w, new_upperLeft[1] - h)
for i in range(numPages):
page = input1.getPage(i)
page.cropBox.upperLeft = new_upperLeft
page.cropBox.lowerRight = new_lowerRight
output.addPage(page)
with open("out.pdf", "wb") as out_f:
output.write(out_f)
Note: in PyPDF2 the origin of coordinates placed in the lower left corner of a page. And the Y-axis is directed from the bottom to up. Not like on the screen. So if you want to get a PDF-coordinate of top edge of your crop area you need to subtract y-coordinate of the top edge of the crop area from the height of the page.