Problem of pages being overwritten while using pytesseract and PyPDF2

Question

I have the following code which aims to turn scanned pdf files into searchable (ctrl+F) files. But for some reason that I cannot understand, the program sometimes (not always) mixes the pages in a way that, for example, it outputs the following: page1, page2, page3, page2, page5... so it writes page2 again on top of page4.

I tried to make the code reproducable.

import pytesseract , PyPDF2 , io , os , tempfile
from pdf2image import convert_from_path
from reportlab.lib.pagesizes import letter
from reportlab.pdfgen import canvas

# set required paths
pdf_folder_path = 'U:/pdf_folder/'
poppler_path = r'C:\poppler-24.02.0\Library\bin'
pytesseract.pytesseract.tesseract_cmd = r'C:\Users\User\AppData\Local\Programs\Tesseract-OCR\tesseract.exe'

# create example pdf file
c = canvas.Canvas("output.pdf", pagesize=letter)
width, height = letter
c.setFont("Helvetica", 100)

for i in range(50):
    c.saveState()
    c.translate(width/2, height/2)
    c.drawCentredString(0, 0, str(i + 1))
    c.restoreState()
    c.showPage()
c.save()

# list all pdfs in the folder (to be made searchable)
directory = os.fsencode(pdf_folder_path)
file_paths = []

for file in os.listdir(directory):
    filename = os.fsdecode(file)
    if filename.endswith('.pdf'): 
        file_paths.append(filename)
        continue
    else:
        continue

print('Files found:')

# convert scanned files into searchable format
for file in file_paths:

    original_pdf = 'U:/Make pdfs searchable/' + file
    searchable_pdf = original_pdf[:-4] + ' (searchable).pdf'

    print('Currently working on: ' + original_pdf)

    with tempfile.TemporaryDirectory() as path:
        images = convert_from_path(original_pdf, poppler_path=poppler_path, output_folder=path)
        pdf_writer = PyPDF2.PdfWriter()
        for image in images:
            page = pytesseract.image_to_pdf_or_hocr(image, extension='pdf', lang='eng')
            pdf = PyPDF2.PdfReader(io.BytesIO(page))
            pdf_writer.add_page(pdf.pages[0])
        with open(searchable_pdf, "wb") as f:
            pdf_writer.write(f)
    print(original_pdf + ' done!')

print('Process complete.')

Any ideas to why this is happening ? Thanks :)

Update 1/3: I have not figured out the cause of the problem, but I created a workaround using PdfMerger() which is however a bit slower.

for file in os.listdir(input_folder):

    if file.endswith(".pdf"):

        original_pdf = input_folder + file
        modified_pdf = output_folder + file[:-4] + ' (searchable).pdf'

        print('Current pdf: ' + file)

        with tempfile.TemporaryDirectory() as path:
            images = convert_from_path(original_pdf, poppler_path=poppler_path, output_folder=path)
            merger = PyPDF2.PdfMerger()
            for count, image in enumerate(images, 1):
                print('Currently on page: ' + str(count))
                page = pytesseract.image_to_pdf_or_hocr(image, extension='pdf', lang='eng')
                pdf = PyPDF2.PdfReader(io.BytesIO(page))
                merger.append(io.BytesIO(page))
            with open(modified_pdf, 'wb') as combined_pdf_file:
                merger.write(combined_pdf_file)

        print(file + ' done!')

print('Process complete.')

Answer 1

It often happens when you take pages directly from the PdfReader object. Let's adjust your code to keep track of the order of the pages before we work with them:

Change this:

for file in file_paths:

    original_pdf = 'U:/Make pdfs searchable/' + file
    searchable_pdf = original_pdf[:-4] + ' (searchable).pdf'

    print('Currently working on: ' + original_pdf)

    with tempfile.TemporaryDirectory() as path:
        images = convert_from_path(original_pdf, poppler_path=poppler_path, output_folder=path)
        pdf_writer = PyPDF2.PdfWriter()
        for image in images:
            page = pytesseract.image_to_pdf_or_hocr(image, extension='pdf', lang='eng')
            pdf = PyPDF2.PdfReader(io.BytesIO(page))
            pdf_writer.add_page(pdf.pages[0])
        with open(searchable_pdf, "wb") as f:
            pdf_writer.write(f)
    print(original_pdf + ' done!')

print('Process complete.')

to this:

# convert scanned files into searchable format
for file in file_paths:
    original_pdf = 'U:/Make pdfs searchable/' + file
    searchable_pdf = original_pdf[:-4] + ' (searchable).pdf'

    print('Currently working on: ' + original_pdf)

    with tempfile.TemporaryDirectory() as path:
        images = convert_from_path(original_pdf, poppler_path=poppler_path, output_folder=path)

        # keeping track of original page numbers
        original_page_numbers = list(range(1, len(images) + 1))

        # initialize dict to store page content
        page_contents = {}

        for i, image in enumerate(images, 1):
            page = pytesseract.image_to_pdf_or_hocr(image, extension='pdf', lang='eng')
            page_contents[i] = page

        # sort pages based on original page numbers
        sorted_pages = [page_contents[i] for i in original_page_numbers]

        # write sorted pages to searchable PDF
        pdf_writer = PyPDF2.PdfWriter()
        for page in sorted_pages:
            pdf = PyPDF2.PdfReader(io.BytesIO(page))
            pdf_writer.add_page(pdf.pages[0])

        with open(searchable_pdf, "wb") as f:
            pdf_writer.write(f)

    print(original_pdf + ' done!')

print('Process complete.')

This should keep your order of pages.