EasyOCR在Python代码中运行正常，转EXE后报错求助

问题详情 #

Python环境下以下EasyOCR文字识别代码可正常运行，但通过auto-py-to-exe转换为EXE文件后出现OpenCV相关报错。

运行正常的Python代码 #

import easyocr
cropped_image_path = 'D:\\Desktop\\screens\\screens\\1.png'
reader = easyocr.Reader(['en'])
easy_results = reader.readtext(cropped_image_path)
print(easy_results)
print("Extracted text is: ", easy_results[0][1])

EXE运行报错信息 #

Neither CUDA nor MPS are available - defaulting to CPU. Note: This module is much faster with a GPU.

Traceback (most recent call last):
File "easyocr_check.py", line 11, in <module>
File "easyocr\\easyocr.py", line 452, in readtext
File "easyocr\\easyocr.py", line 321, in detect
File "easyocr\\detection.py", line 95, in get_textbox
File "easyocr\\detection.py", line 55, in test_net
File "easyocr\\craft_utils.py", line 236, in getDetBoxes
File "easyocr\\craft_utils.py", line 31, in getDetBoxes_core
cv2.error: Unknown C++ exception from OpenCV code
[8260] Failed to execute script 'easyocr_check' due to unhandled exception!

使用版本 #

• Python v3.9
• easyocr v1.7.0
• opencv-python v4.7.0.72
• auto-py-to-exe v2.36.0

Python环境预期输出 #

[([[23, 5], [59, 5], [59, 19], [23, 19]], '21.23', 0.9985201957308608)]
Extracted text is:  21.23

解决方案 #

• 降级OpenCV版本
  高版本opencv-python打包后易出现兼容性问题，降级到4.5.x稳定版本，执行命令：

  pip install opencv-python==4.5.5.62
  

  卸载原版本后重新安装，再重新打包。

• 手动指定EasyOCR模型路径
  打包时可能未正确包含模型文件，手动下载模型到本地目录，修改代码指定模型存储路径：

  import easyocr
  cropped_image_path = 'D:\\Desktop\\screens\\screens\\1.png'
  # 替换为你本地的模型文件夹路径
  reader = easyocr.Reader(['en'], model_storage_directory='./easyocr_models')
  easy_results = reader.readtext(cropped_image_path)
  print(easy_results)
  print("Extracted text is: ", easy_results[0][1])
  

  打包时将easyocr_models文件夹与EXE放在同一目录，或在auto-py-to-exe的Additional Files中添加该文件夹。

• 调整auto-py-to-exe打包参数

  1. 选择One Directory模式打包（不选One File），依赖文件会全部存放在同一文件夹，减少路径解析问题。
  2. 在Additional Files中添加EasyOCR默认模型目录（通常为C:\Users\<你的用户名>\.EasyOCR\model），确保打包后模型文件能被EXE访问。

• 修正图像路径问题
  打包后EXE的工作路径可能与Python环境不同，避免硬编码绝对路径，改为让用户选择图像文件：

  import easyocr
  from tkinter import filedialog
  
  cropped_image_path = filedialog.askopenfilename(filetypes=[("Image Files", "*.png;*.jpg")])
  reader = easyocr.Reader(['en'])
  easy_results = reader.readtext(cropped_image_path)
  print(easy_results)
  print("Extracted text is: ", easy_results[0][1])
  

内容的提问来源于stack exchange，提问作者Test One