Hi,

thank you for your question.

You can find a sample for a pipeline that uses doctr instead of Tesseract in the Space demo that unfortunately doesn't not work but for a different reason.

You will need to build a pipeline by yourself but you can copy the code from build_analyzer(cfg) in deepdoctection.analyzer.dd and replace the part where Tesseract is invoked (line 193-198) with

# word detector
det = dd.DoctrTextlineDetector()
  
# text recognizer
rec = dd.DoctrTextRecognizer()

d_layout_text = dd.ImageLayoutService(det, to_image=True, crop_image=True)

pipe_component_list.append(d_layout_text)
 
d_text = dd.TextExtractionService(rec, extract_from_roi="WORD")
pipe_component_list.append(d_text)

Feel free to ask if there are any further questions.