stringandstickytape / RegulatedNoise

20 stars 16 forks source link

several optimizations and some bugs fixed #57

Closed Duke-Jones closed 9 years ago

Duke-Jones commented 9 years ago

03.02.15: comment updated - see below

Hi, great tool !

I was only a little bit unsatisfied with the recognition rate and so I testet different brightness and contrast settings with the original TesseractOCR command line client. After that I tested the different "traineddata" files as well. It turned out that the "big.traineddata" from EliteOCR results in the best recognition rate. I think it's very good now. Additionally I implemented a dynamic factor intstead a constant value for accepting the Leventhstein words. So longer words allowing more incorrect characters. But so we can consider that they've also more correct charcters, too. It works pretty good.

While testing I was able to fix a bug and I' added a few other things. I hope you like it.

Duke

PS: sorry for my bad english - I hope you can understand me ;-)

03.02.15 - added something:

1.8.4-0.0.1

Donavan commented 9 years ago

Pulling this puts the project into an uncompilable state unless you define DukeJones.

Duke-Jones commented 9 years ago

Oh, I'm sorry. I fixed this.

DrAvEn83 commented 9 years ago

hi Duke Nice job

but can you add a butten for Delet 1 Commodities, and LS Filter add

im happy too see you work good