from pdf2image import convert_from_path

def get_page_image(  # pdf 按页切成图片
        hash_value: str
) -> Dict[int, str]:
    pdf_path = os.path.join(config.PDF_DIR, f‘{hash_value}.pdf‘).replace(‘\\‘, ‘/‘)
    img_pages = convert_from_path(pdf_path)
    res = {}
    for page, img in enumerate(img_pages):
        arr = np.array(img)
        basename = f‘{hash_value}_{page:03d}.jpg‘
        image_path = os.path.join(config.PDF_IMAGE_DIR, basename)
        img.save(image_path)
        res[page] = image_path
    return res

pdfpath = os.path.join(config.PDF_DIR, paper_id + ".pdf")
xmlpath = os.path.join(config.PDF_XML_DIR, paper_id + "_{}_pdfminer.xml".format(page))
if not os.path.exists(xmlpath):
    os.system("pdf2txt.py -o {} -p {} -t xml {}".format(xmlpath, page + 1, pdfpath))