Thanks for suggestions. You are right that I'm reffering to api.GetUTF8Text() call, it is my bottleneck. I was not aware that there is some fast and best models in tesseract 4.0, I will give it a try. So far I used just lang=eng or osd. For me it is suspicious that the calls are getting longer during the time. Or to be more precise, first 10-15 calls are up to 500ms and latter ones rise above 1s... SetSourceResolution outside of the loop gives no change unfortunately.
BR Jan Dne úterý 29. ledna 2019 18:08:49 UTC+1 Lorenzo Blz napsal(a): > > > First double check if the Pi is not throttling due to overheating or lack > of USB power. This may cause the slowdown. > > Usually 30/50 px of text height is fine. IF the problem is tesseract, try > to use the fast model (or "normal" if using best). I assume you are using > the 4.x release. > > Try tesseract -v to see if you are using all the available CPU > optimizations. > > Try to move the SetSourceResolution outside the loop and see if it changes > something (MAYBE it may invalidate some caches or something). > > The time you are referring to is one single api.GetUTF8Text() call, > correct? > > > Lorenzo > > > Il giorno mar 29 gen 2019 alle ore 17:48 Jan Pohanka <[email protected] > <javascript:>> ha scritto: > >> Hello, >> >> I'm making a simple device used to recognize numbers on pictures taken by >> a webcam. All is running on raspberry pi 3. >> Everything is like following simple loop (in python for simplicity, but >> using C++ api it is the same), images are preprocessed to black and white >> >> api = PyTessBaseAPI(psm=tesserocr.PSM.SINGLE_WORD) >> >> for im in images: >> api.SetImage(im) >> api.SetSourceResolution(70) >> ot = api.GetUTF8Text() >> >> api.End() >> >> >> My problem is that api.GetUTF8Text() call is quite slow and more over it >> is getting slower and slower over time. Is there any options how to make >> recognition faster? I have tried to resize the image to around 50x10px. The >> times starts on around 300ms but then goes up to above 1s which is too slow >> for me. I tried both legacy and LSTM algorithms, but they are similar. >> >> best regards >> Jan >> >> -- >> You received this message because you are subscribed to the Google Groups >> "tesseract-ocr" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected] <javascript:>. >> To post to this group, send email to [email protected] >> <javascript:>. >> Visit this group at https://groups.google.com/group/tesseract-ocr. >> To view this discussion on the web visit >> https://groups.google.com/d/msgid/tesseract-ocr/a53b4b25-97e3-47dc-823a-cbb219225eed%40googlegroups.com >> >> <https://groups.google.com/d/msgid/tesseract-ocr/a53b4b25-97e3-47dc-823a-cbb219225eed%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> For more options, visit https://groups.google.com/d/optout. >> > -- You received this message because you are subscribed to the Google Groups "tesseract-ocr" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at https://groups.google.com/group/tesseract-ocr. To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/baa59c86-b002-4607-8dda-16835cd3ea73%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
