调用开源LaTex OCR的一个简单实现（很实用）

注意事项

1. 根据以下链接进行配置，具体的使用情况可以看我的分享【免费开源的LaTex OCR软件分享（latex-ocr和texify使用分享）】

LaTeX-OCR: https://github.com/lukas-blecher/LaTeX-OCR

texify: https://github.com/VikParuchuri/texify

2. 配置环境可能存在某些库版本太新，导致安装出现异常;
3. 下面剪贴板的路径修改为自己存在的路径；
4. 使用F5刷新界面来获取新的识别内容；

texify调用

# 调用命令
streamlit run .\texify_app.py

import os
from PIL import Image, ImageGrab
import streamlit as st
from texify.inference import batch_inference
from texify.model.model import load_model
from texify.model.processor import load_processor

# how to use
# streamlit run .\texify_app.py

# 注释掉的部分可以调用文件上传，但感觉对我不是很实用

# if __name__ == '__main__':
#     streamlit.set_page_config(page_title='Texify OCR')
#     streamlit.title('Texify OCR')
#     streamlit.markdown(
#         'Convert images of equations to corresponding LaTeX code.\n\nThis is based on the `texify` module.')

#     uploaded_file = streamlit.file_uploader(
#         'Upload an image an equation',
#         type=['png', 'jpg'],
#     )

#     if uploaded_file is not None:
#         image = Image.open(uploaded_file)
#         streamlit.image(image)
#     else:
#         streamlit.text('\n')

#     if streamlit.button('Convert'):
#         if uploaded_file is not None and image is not None:
#             with streamlit.spinner('Computing'):
#                 try:
#                     latex_code = batch_inference([image], model, processor)
#                     streamlit.code(latex_code, language='latex')
#                     streamlit.markdown(f'$\\displaystyle {latex_code}$')
#                 except Exception as e:
#                     streamlit.error(e)
#         else:
#             streamlit.error('Please upload an image.')


if __name__ == '__main__':
    st.set_page_config(page_title='LaTeX-OCR')
    st.title('LaTeX OCR')
    st.markdown('Convert images of equations to corresponding LaTeX code.')
    image_container = st.empty()
    im = ImageGrab.grabclipboard()
    import time
    time.sleep(0.1)
    
    flie_path = "D:\\Miniconda\\envs\\texocr\\doc\\clipboard\\grab_clipboard.jpg"
    if isinstance(im, Image.Image):
        print("Image: size : %s, mode: %s" % (im.size, im.mode))
        image_container.image(im)
        im.save(flie_path)
    else:
        print("clipboard is empty")
    
    if os.path.exists(flie_path):
        if st.button('Convert'):
            with st.spinner('Computing'):
                try:
                    model = load_model()
                    processor = load_processor()

                    image = Image.open(flie_path)
                    latex_code = batch_inference([image], model, processor)
                    # print(latex_code)
                    st.code(latex_code[0], language='latex')
                    st.markdown(latex_code[0])
                except Exception as e:
                    st.error('Failed Convert!')
    else:
        st.info('Please copy an image to your clipboard first, then click "Use Clipboard"')

latex ocr调用

# 运行命令
streamlit run .\ocrtex_app.py

import os
from PIL import Image, ImageGrab
import streamlit as st
from pix2tex.cli import LatexOCR
from munch import Munch

args = Munch({'config': 'settings/config.yaml',
              'checkpoint': 'checkpoints/weights.pth',
              'no_resize': False})

if __name__ == '__main__':
    st.set_page_config(page_title='LaTeX-OCR')
    st.title('LaTeX OCR')
    st.markdown('Convert images of equations to corresponding LaTeX code.')
    image_container = st.empty()
    im = ImageGrab.grabclipboard()
    import time
    time.sleep(0.1)
    
    flie_path = "D:\\Miniconda\\envs\\texocr\\doc\\clipboard\\grab_clipboard.jpg"
    if isinstance(im, Image.Image):
        print("Image: size : %s, mode: %s" % (im.size, im.mode))
        image_container.image(im)
        im.save(flie_path)
    else:
        print("clipboard is empty")
    
    if os.path.exists(flie_path):
        if st.button('Convert'):
            with st.spinner('Computing'):
                try:
                    model = LatexOCR(args)
                    image = Image.open(flie_path)
                    latex_code = model(image)
                    print(latex_code)
                    cp_content = '$$ ' + latex_code + ' $$'
                    st.code(cp_content, language='latex')
                    st.markdown(f'$\\displaystyle {latex_code}$')
                except Exception as e:
                    st.error('Failed Convert!')
    else:
        st.info('Please copy an image to your clipboard first, then click "Use Clipboard"')

文本OCR

图片示例如下，背景简单没问题，稍微复杂的背景可能不太行，一般是建议使用微信，几乎大家都有，使用简单。

https://github.com/deanmalmgren/textract

按照上述同样#1#2的原理可以很容易实现下面效果：

注意：需要单独下载支持中文的语言包，还要安装支持的库，这个已经被作者封装好exe，直接安装就行。

import os
import textract
from PIL import Image, ImageGrab
import streamlit as st

if __name__ == '__main__':
    st.set_page_config(page_title='Text-OCR')
    st.title('Text OCR')
    st.markdown('Convert images of equations to corresponding LaTeX code.')
    image_container = st.empty()
    im = ImageGrab.grabclipboard()
    import time
    time.sleep(0.1)
    
    flie_path = "D:\\Miniconda\\envs\\texocr\\doc\\clipboard\\grab_clipboard.jpg"
    if isinstance(im, Image.Image):
        print("Image: size : %s, mode: %s" % (im.size, im.mode))
        image_container.image(im)
        im.save(flie_path)
    else:
        print("clipboard is empty")
    
    if os.path.exists(flie_path):
        if st.button('Convert'):
            with st.spinner('Computing'):
                try:
                    full_text = textract.process(
                                    flie_path,
                                    method='tesseract',
                                    language='chi_sim',
                                ).decode('utf-8')
                    # print(full_text)
                    st.code(full_text, language='txt')
                    st.markdown(full_text)
                except Exception as e:
                    st.error('Failed Convert!')
    else:
        st.info('Please copy an image to your clipboard first, then click "Use Clipboard"')