Quick and dirty implementation of a text and bounding box extraction from PDFs using pdfminer.six==20191110

pdf_mine_with_boxes.py

from collections import defaultdict

import fire
from pdfminer.converter import PDFPageAggregator
from pdfminer.layout import LAParams
from pdfminer.layout import LTChar
from pdfminer.layout import LTTextBox
from pdfminer.layout import LTTextLine
from pdfminer.pdfinterp import PDFPageInterpreter
from pdfminer.pdfinterp import PDFResourceManager
from pdfminer.pdfpage import PDFPage
from rich import print as rprint


def get_char_bboxes(pdf_path: str):
    fp = open(pdf_path, "rb")
    rsrcmgr = PDFResourceManager()
    laparams = LAParams()
    device = PDFPageAggregator(rsrcmgr, laparams=laparams)
    interpreter = PDFPageInterpreter(rsrcmgr, device)
    pages = PDFPage.get_pages(fp)

    per_page_tboxes = defaultdict(list)
    for i, page in enumerate(pages):
        print("Processing next page...")
        interpreter.process_page(page)
        layout = device.get_result()
        for lobj in layout:
            if isinstance(lobj, LTTextBox):
                for tline in lobj:
                    for tchar in tline:
                        if isinstance(tchar, LTChar):
                            # x0, y0, x1, y1 = tchar.bbox[0], tchar.bbox[1], tchar.bbox[2], tchar.bbox[3]
                            # text = tchar.get_text()
                            per_page_tboxes[i].append((tchar.bbox, tchar.get_text()))

    return per_page_tboxes


def textract(pdf_path: str):
    per_page_bboxes = get_char_bboxes(pdf_path)

    for p_idx, page_bboxes in per_page_bboxes.items():
        # 1. Group by vertical position
        groups = defaultdict(list)
        for tbox in page_bboxes:
            coords, _ = tbox
            _, y0, _, _ = coords
            groups[y0].append(tbox)

        # 2. Split by space and get the enclosing box
        for i, (y0, tbox_list) in enumerate(groups.items()):
            # Group in words
            words = []
            wboxes = []
            bb = []
            w = ""
            for tbox in tbox_list:
                try:
                    coords, c = tbox
                    if c != " ":
                        w += c
                        bb.append(coords)
                    else:
                        if len(bb):
                            words.append(w)
                            if len(bb) > 1:
                                bb_s, bb_f = bb[0], bb[-1]
                                wboxes.append((bb_s[0], bb_s[1], bb_f[2], bb_f[3]))
                            else:
                                wboxes.append((bb[0]))
                            w = ""
                            bb = []
                except Exception as e:
                    rprint(f"[red] Error: {e}[/red]")
                    rprint(f"[red] w: {w} | bb: {bb}[/red]")

            tboxes = list(zip(words, wboxes))
            rprint(f"PAGE {p_idx} - LINE: {i}")
            rprint(tboxes)


if __name__ == "__main__":
    # Requires:
    # fire==0.5.0
    # pdfminer.six==20191110
    # rich==13.6.0
    fire.Fire({"pdf": textract})