Paddleocr integration drops accuracy

Hello, good day to everyone
I would like to integrate a library to perform ocr on images. I manage to integrate paddleOCR but I notice that the accuracy drops a lot when I compare the same model in Google Colab. Any idea or suggestion.


  - conda-forge

  - python=3.7.5
  - pip=20.1
  - python-Levenshtein=0.12.2
  - pip:
      - rpaframework==13.0.0 #
      - paddlepaddle==2.2.2
      - paddleocr==2.4.0
      - opencv-python==

from paddleocr import PaddleOCR

def get_word(img_path):
    ocr = PaddleOCR(use_angle_cls=True, lang='en', use_gpu=False)
    result = ocr.ocr(img_path, det=False, cls=True)

Thanks a lot

Hi there, maybe you find our rpaframework-recognition library enough for your use case. (if you plan to use it in your Python robot)

Some robot examples from our portal using it:

AI related articles/keywords in docs:

Hello, thanks for the answer. I was investigating and within the OCR libraries the most accurate was PaddleOCR. I need to get the text of an image and others like tesseract or EasyOCR have less precision.

1 Like