How pycharm packages OCR correctly and makes the packaged exe as small as possible

pack ocr I encountered many problems in the process of , The whole process involves path modification and environment variables . I believe many people will encounter the same problems as me , I try to write down all the problems I encounter in the process of packaging . I hope it will help you .

At first, my development environment was set up like this

Chose the third one   System environment   , It turns out that  pyinstaller When the packaging , Put all the libraries in the system Including my py Libraries not used by the main program All packed in , The result is naturally a large file .

  To solve this problem , I choose the first one instead   Virtualenv  Environment  , A virtual environment is established under the root directory of the main program , But when you click ok , But prompted me to say ： Permission denied .

The essence of the problem is , You installed python when , Choose a fool installation ,python Installed under your personal name . So now use this python Interpreter , The system thinks that you are in your own name , Not the whole computer , So I think I don't have this permission , rejected .

The solution is ： It is recommended to reinstall python. I downloaded.

3.9.5 Version of , Double click installation .

  Below Add python 3.9 to PATH hit √, Then click on Customize  installation . Then enter the next interface , Remember right   for  all  user  hit √, Then click next installation step by step Until the end .

After this , Yours python , It's installed in Under this path . And the permission is not restricted .

In this way, you can successfully set up the virtual environment .

Scanning class pdf turn word When writing , Basically, you need ocr Tools .

Install in advance ocr, The fool installation will be installed in C disc .

my py The main program , The modules used are ：

among os yes python Self contained , No installation required , Others passed pycharm Below Terminal Installed ：

Terminal, adopt pip install xxx Install the modules you need

pip  install  PIL Will make mistakes , To be changed to pip  install  image Can .

Run times No module named 'frontend', By additional installation pip install PyMuPDF solve ;

newspaper Error: No module named 'exceptions', adopt pip install python-docx solve ;

Then the following error message appears ,

This problem can be solved by pip install PyMuPDF==1.19.0 solve , Before that PyMuPDF Version of is too low , It's not pageCount Of this property .

（（

By the way , Prompt during operation DeprecationWarning: ANTIALIAS is deprecated and will be removed in Pillow 10 (2023-07-01). Use Resampling.LANCZOS instead.

say ANTIALIAS Has been deprecated , When I change the pixels of the picture , Just used Image.ANTIALIAS, Source code ：

I simply changed the above sentence into

It turns out that the identified error rate is even lower . This is an empirical statement .

））

Continue the topic of packaging ：

py Runtime , It is to find site-packages\pytesseract\pytesseract.py file .

There are two possibilities ： If you are setting conda As an interpreter rather than a virtual environment , That system will read pytesseract.py When you file , Read the specific location in D:\ProgramData\Anaconda3\Lib\site-packages\pytesseract\pytesseract.py This .

（ The system will also pack this when you pack py. therefore , If your location guide needs to be modified , Just change this py.）

If you set up a virtual environment ,venv Build in the same root directory of your main program , When the program is running or packaged , The file found by the system , It is under the root directory of the main program py file . The specific location is

（py The root directory of the main program ）+  \venv\Lib\site-packages\pytesseract\pytesseract.py

And then through pytesseract.py Inside

This code , Assigned a value to tesseract The location of , Let the system pass the environment variables you set before , find ocr Tools .

If you don't set the environment variables , Then change the above sentence into

Let the system go directly to the tools you installed .

Follow the steps above , Packed exe , It can run on your computer , But if you go to someone else's computer, it won't work . Because other people's computers are not like yours C There are tesseract The tool .

therefore , Most of the above online modification methods , It can only be used on your computer , pack exe If you go to someone else's computer to run, you will report an error

Say you didn't install or didn't set up the virtual environment after installation .

resolvent , Go to C Handlebar ocr Tool copy out , Put it in your py Under the root directory of the main program .

then pytesseract.py in , It corresponds to writing , Lead the path to this under the root directory ocr,

  It looks like , Packed exe, The root directory is placed from C Copy from disk  Tesseract-OCR Folder , Send it to others together and you can use it .

There is another point to pay special attention to ： stay pyinstaller  -F  XX.py Before , Be sure to pip  install  pyinstaller install pyinstaller. Otherwise, it can also generate exe, But generated exe The file will be many times larger .

After all this fuss , Finally, I packed the file from the previous 347M, Shrink to the present 66M, Shrunk 5 Twice as many .

Then we think , Can you put this auxiliary folder  Tesseract-OCR Pack it together. It just came out exe Well ？

Please read my previous article first , The principle must be understood , Otherwise, you will see the silly circle .

python How to put the auxiliary files needed for program running , Pack together into exe in _py617 The blog of -CSDN Blog

Next, let's talk about my solution to this problem ：

pytesseract.py Make such changes in ：

  The code is as follows ：

import sys   
import os
def resource_path(path_a):
    if getattr(sys, 'frozen', False): # whether Bundle Resource
        base_path = sys._MEIPASS
    else:
        base_path = os.path.abspath(".")
    return os.path.join(base_path, path_a)
fuzhu_filename = resource_path(os.path.join("",))

tesseract_cmd = os.path.abspath(os.path.join(fuzhu_filename,'Tesseract-OCR','tesseract.exe'))

Then open the first packaging exe when , Generated under the same root directory spec file . Make the following modifications ： 

  It means when you want to package files tpdf_docx.py , The address is xxx, And then pack it up py When you file , together with Tesseract-OCR To pack together , After packing , The name is also Tesseract-OCR.

Last , stay pycharm Of test.py Enter pyinstaller tpdf_docx.spec. Press Enter Key wait , The final file is generated in dist in . Generated for the first time before exe Will be replaced , If you want that first generated exe, It's just pyinstaller tpdf_docx.spec Before , Just copy it first .

This problem needs to be tried constantly , There will always be problems in the middle , But as long as you are good at thinking , Believe that the process is tortuous , The future is bright .