Don't install this module without reading the instructions!!

The best OCR engine

In my opinion, EasyOCR is the best OCR engine out there.

The results are much better than Tesseract's OCR, but it is much slower.

And, unfortunately, it is a bit tricky to install.

Example - Bluestacks [adb/hwnd]

  <a href="https://www.youtube.com/watch?v=hDRwmzbrvaY">

     <img src="https://img.youtube.com/vi/hDRwmzbrvaY/0.jpg" style="width:100%;">

  </a>

Step 1

If you have a GPU and Windows (by the way, this module is only for Windows), you should install torch and torchvision first to get a good speed up

*From https://github.com/JaidedAI/EasyOCR

Note 1: For Windows, please install torch and torchvision first by following the official instructions here https://pytorch.org. On the pytorch website, be sure to select the right CUDA version you have. If you intend to run on CPU mode only, select CUDA = None.*

Step 2

pip install easyocr_window_scanner

Step 3

EasyOCR installs the headless (light) version of opencv that doesn't even have cv2.imshow, so

EasyOCR2Df.scan_window_with_adb / EasyOCR2Df.scan_window_with_hwnd don't work!

To fix this problem, do this:

pip uninstall opencv-python

pip uninstall opencv-python-headless

pip install --upgrade --force-reinstall opencv-python==4.5.5.62

Step 4

You might get the following error when you import the module:

cv2.error: Unknown C++ exception from OpenCV code.

It seems to be a very common issue with EasyOCR

https://github.com/opencv/opencv-python/issues/602#issuecomment-1070290110

To fix this problem, you have to open:

"C:\Users\YOURUSERNAME\anaconda3\envs\YOURENV\Lib\site-packages\easyocr\craft_utils.py"

and change:

nLabels, labels, stats, centroids = cv2.connectedComponentsWithStats(text_score_comb.astype(np.uint8), connectivity=4)

to:

nLabels, labels, stats, centroids = cv2.connectedComponentsWithStats(text_score_comb.astype(np.uint8))

Step 5

#We are finally ready to go



from easyocr_window_scanner import EasyOCR2Df



#create an instance, you can pass any argument you would pass to the EasyOCR module

easyo = EasyOCR2Df(["en"])



#Taking screenshots using adb, and do OCR on the fly

easyo.scan_window_with_adb(adb_path=r"C:\ProgramData\adb\adb.exe",adb_serial='localhost:5555',quit_key='q')



#Taking screenshots using the windows handle, and do OCR on the fly

easyo.scan_window_with_hwnd(hwnd=None, regular_expression='[bB]lue[sS]tacks.*', quit_key="q")



#If you just want to have the OCR results as a DataFrame, use:

easyo.start_ocr_to_df(r'https://i.ytimg.com/vi/fa82Qpw6lyE/hqdefault.jpg').get_results_df()

Out[5]:

                               text      conf  ...  middle_x  middle_y

0     No stop signs, speedin' limit  0.842704  ...       148        90

1       Nobody's gonna slow me down  0.744984  ...       159       112

2       Like a wheel, gonna spin it  0.966272  ...       141       136

3     Nobody's gonna mess me 'round  0.994361  ...       165       157

4               Satan! Paid my dues  0.708869  ...       153       180

5         Playin' in a rockin' band  0.979712  ...       133       202

6                  Mama' Look at me  0.559171  ...       147       224

7  Tm on my way t0 the promise land  0.703882  ...       174       247

8                               Hey  0.999969  ...        53       179

9                               Hey  0.999027  ...        52       224

[10 rows x 14 columns]



#If you want to have the image of the OCR results without showing it, use:

easyo.start_ocr_to_df(r'https://i.ytimg.com/vi/fa82Qpw6lyE/hqdefault.jpg').draw_easyocr_results().get_results_picture()

Out[7]:

array([[[0, 0, 0],

        [0, 0, 0],

        [0, 0, 0],

        ...,

        [0, 0, 0],

        [0, 0, 0],

        [0, 0, 0]],

        ...