Heh. It's an old issue. For 100% accuracy, you must use a digit-only language model. But there is no such thing. Besides trivial perceptron shows good results on digits recognition. суббота, 30 января 2021 г. в 18:41:13 UTC+3, Benek:
> Hello! I'm trying to read some digits and I thought it was a rather simple > task yet still I can't receive satisfying results. So my first question is: > is it possible to get 100% accuracy when reading some standardized input? > Or there will be always some errors when reading? > > Here are some sample inputs that I wanted to read: > The digits that are being misread are: > on the photo t3 > 5.1 is read as 9.1 > on the photo t4 > : 10.2 is read as 610.2 > > I'm using: > > tesseract/4.1.1 > > config : > > oem: 3, > > psm: 11, > > tessedit_char_whitelist: "0123456789.", > > load_system_dawg: false, > > load_freq_dawg: false, > > The images have 2700x2100 resolution. > > The 999 on the left are markers that I added to be able to recognize which > line belongs to which output text and they are always read correctly. > I tried experimenting with some different image preprocessing techniques > like blur, median, changing the size of the image etc. > > Do you have any other tips that could lead to better reading accuracy? > Thanks in advance for any help! -- 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 tesseract-ocr+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/14120ee4-9d78-48ff-b088-d3fa503aefe5n%40googlegroups.com.
