Screen-OCR Documentation Need

[brighten6363]

Hello?
I've encountered screen-ocr[winrt] and I'm going to try it.
like.. lang=language, psm=6..etc.. used by tesseract.. How do I set the config parameter with screen-ocr??
I want to control parameters that psm, lang... ect.

If the block area is recognized using ocr_reader.read_screen(area),
the string is not output continuously.
Depending on the +/- number value or color of the letter, the order is not continuous and the order is mixed up.

I want to solve this problem.
Like using config parameter in tesseract..

What is the definition of kwargs in create_quality_reader(cls, **kwargs)?
I'd appreciate it if you could show me an example of how to use it.

I'm asking because I can't figure it out even if I look at the example.
thank U.

[wolfmanstout]

You can pass any of the following keyword arguments to create_quality_reader:

screen-ocr/screen_ocr/_screen_ocr.py

Lines 93 to 102 in ef2dcd1

	backend: Union[str, _base.OcrBackend],
	tesseract_data_path=None,
	tesseract_command=None,
	threshold_function="local_otsu",
	threshold_block_size=41,
	correction_block_size=31,
	convert_grayscale=True,
	shift_channels=True,
	debug_image_callback=None,
	language_tag=None,

screen-ocr/screen_ocr/_screen_ocr.py

Lines 166 to 174 in ef2dcd1

	backend: _base.OcrBackend,
	margin: int = 0,
	resize_factor: int = 1,
	resize_method=None, # Pillow resize method
	debug_image_callback: Optional[Callable[[str, Any], None]] = None,
	confidence_threshold: float = 0.75,
	radius: int = 200, # screenshot "radius"
	search_radius: int = 125,
	homophones: Optional[Mapping[str, Iterable[str]]] = None,

Please note that some of the arguments in the first set only apply to one backend or another. It is unclear from your question whether you plan to use winrt or tesseract -- you will need to choose one or the other, and the former only works on Windows. If you do want to use tesseract, currently there is no way to directly override the config that gets passed to pytesseract, although it wouldn't be too difficult to plumb this through if that's what you need. Here is where the call to pytesseract actually happens:

screen-ocr/screen_ocr/_tesseract.py

Lines 45 to 46 in ef2dcd1

	results = pytesseract.image_to_data(
	image, config=tessdata_dir_config, output_type=pytesseract.Output.DATAFRAME

[brighten6363]

Thank you for your answer.
My questions, It's all about winrt.
In tesseract, whether to treat images as blocks, lines, or columns can be controlled by delivering psm values, but I wonder how it is possible in winrt.
Thank you~

[wolfmanstout]

Got it. WinRT has very minimal configuration options: https://learn.microsoft.com/en-us/uwp/api/windows.media.ocr.ocrengine?view=winrt-22621

[brighten6363]

Thanks for your help.
Then, I don't think it would be easy to extract text from the image block in order. If It use winrt libarary..

But I don't understand why it's not output in order. For example, a table consisting of rows and columns with +/- numbers, winrt printout later some columns are ahead of other numbers. The same image on the tesseract with block config parameter, it will be printout in order..

[wolfmanstout]

I'm assuming you are referring to the ordering in the ScreenContents.as_string function. This is just a convenience function that concatenates in the original order that the OCR library produced:

screen-ocr/screen_ocr/_screen_ocr.py

Lines 440 to 448 in ef2dcd1

	def as_string(self) -> str:
	"""Return the contents formatted as a string."""
	lines = []
	for line in self.result.lines:
	words = []
	for word in line.words:
	words.append(word.text)
	lines.append(" ".join(words) + "\n")
	return "".join(lines)

If you want to do your own order, you are welcome to check the bounds of the OcrLines and concatenate these however you want, using a separate function.

[brighten6363]

Thanks. :-)
Very similar to what I want. I'll test it in many ways.