INNER CODE UNIT · Python

find_caption_block

gokayfem/awesome-vlm-architectures · scripts/extract_architecture_figures.py:52

def find_caption_block(page: fitz.Page, figure: str) -> tuple[fitz.Rect, str]:
    wanted = normalize_figure(figure)
    patterns = [
        re.compile(rf"(?:Figure|Fig\.)\s*{re.escape(wanted)}\b", re.I),
        re.compile(
            rf"(?:Figure|Fig\.)\s*{re.escape(wanted.rstrip('abcdefghijklmnopqrstuvwxyz'))}\b",
            re.I,
        ),
    ]
    matches: list[tuple[fitz.Rect, str]] = []
    for block in page.get_text("blocks", sort=True):
        text = re.sub(r"\s+", " ", block[4]).strip()
        if any(pattern.search(text) for pattern in patterns):
            matches.append((fitz.Rect(block[:4]), text))
    if not matches:
        raise ValueError(
            f"caption for {figure} not found on PDF page {page.number + 1}"
        )

View source record →