使用大模型处理文档时，需要对二进制格式的文档进转解析提取文字和图片，本文使用 pymupdf 开源库，对 PDF 进行解析提取文字和图片。

安装依赖

首先安装 pymupdf 依赖

pymupdf4llm==0.0.17
pymupdf==1.24.10
apscheduler==3.10.4

PDF 转 Markdown

pymupdf 将 PDF 转换为 markdown，传入 pdf 路径进行处理。

def  convert_to_md(pdf_path):
    return pymupdf4llm.to_markdown(pdf_path)

PDF 中提取图片

传入 PDF 路径，保存图片，并返回路径列表。

def extract_image(pdf_path):
    doc = pymupdf.open(pdf_path) # open a document
    image_path_list = []
    rad_prefix = ''.join(random.choices(string.ascii_letters, k=5))
    Path(f"images/{rad_prefix}").mkdir(parents=True)
    for page_index in range(len(doc)): # iterate over pdf pages
        page = doc[page_index] # get the page
        image_list = page.get_images()

        # print the number of images found on the page
        if image_list:
            print(f"Found {len(image_list)} images on page {page_index}")
        else:
            print("No images found on page", page_index)
        
        for image_index, img in enumerate(image_list, start=1): # enumerate the image list
            xref = img[0] # get the XREF of the image
            pix = pymupdf.Pixmap(doc, xref) # create a Pixmap

            if pix.n - pix.alpha > 3: # CMYK: convert to RGB first
                pix = pymupdf.Pixmap(pymupdf.csRGB, pix)
            
            image_path = Path("images/%s/page_%s-image_%s.png" % (rad_prefix,page_index, image_index)).absolute()
            image_path_list.append(image_path)
            pix.pil_save(image_path) # save the image as png
            pix = None
    return image_path_list

可视化

通过 Gradio 可视化 PDF 转换。

import gradio as gr
import pdf
def covnert_pdf(x):
    print(x)
    md_content = pdf.convert_to_md(x)
    image_list = pdf.extract_image(x)
    return [md_content , image_list]


with gr.Blocks() as demo:
    gr.Markdown(
        value="# PDF 提取"
    )
    with gr.Tab("Markdown"):
        md = gr.Markdown(
             height=500
        )
    with gr.Tab("Images"):
        gallery = gr.Gallery(columns=4,  object_fit="None", height="auto", show_download_button=True)
    with gr.Row():
        go_btn = gr.UploadButton()
        go_btn.upload(covnert_pdf, go_btn, [md, gallery])

from apscheduler.schedulers.background import BackgroundScheduler


demo.launch()

总结

pymupdf 转 markdown 的效果还是不错的，可以到 Modelscope 的空间进行体验。https://modelscope.cn/studios/model1001/pdf_converter