The title is wrong – it actually uses Tesseract to do the OCR.
Call me negative, but I don't see the big leap here, it's just a simple demo to recognize text in an almost perfect input picture. It uses the Emgu CV, self description: "Emgu CV is a cross platform .Net wrapper to the OpenCV image processing library.".
I did this back in 2012, using Tesseract and OpenCV to scan letters from screenshots. I did it in python, though, but it was relatively straightforward. I actually had to create my own "language" to help with the processing, bringing it from the low 90s to ~98% accuracy.
It's perfect from an OCR point of view:
- Perfectly upright, no visible rotation
- Sharp, not remotely blurry
- No visible brightness/color gradient (as is often the case with pictures taken by mobile phones.)
It contains umlauts (and i'm not sure whether the tesseract training data for german includes those, but i'm quite certain it does), but that doesnt disqualify the image.
I was just saying hey, it's not using openCV but tesseract, so the title is wrong.. and i stand by that statement. I appreciate the work you do and i was exclusively referring to the title here. have a good day ;)
This looks cool. I didn't try the examples but that receipt example is what I really need.
I keep track of how much money do I spend on what things by going through receipts collected over a month. While the information is useful it's too much manual work. I wish I could just scan or take a picture of my receipts and extract text for further processing.
You can try https://www.expensify.com/ .. they have a really nice App and they are nice people as well .. i read their startup story once and it was really interesting
This is interesting because I have a food project on ice, that got stuck on OCR'ing receipt's. There are proprietary scanners that are OK, but not perfect. There are still a lot of progress to be done is in this area!
So, I checked the code from Nuget's GitHub pages and at the first sight it seems more complex which doesn't have to be 'bad' but its simply more demanding than my small example.
Of course, it also seems that you can do more sophisticated things with this Nuget.
Call me negative, but I don't see the big leap here, it's just a simple demo to recognize text in an almost perfect input picture. It uses the Emgu CV, self description: "Emgu CV is a cross platform .Net wrapper to the OpenCV image processing library.".