Chi_sim under the currently used tess4j, online. There are about 50 MB traineddata, the recognition rate is very low, to their own training and have no time, the company also could not arrange the manpower, have a high recognition rate, about 90%, can apply for to make the company purchased,
Don't consider baidu, ali, tencent's online identification service, because the customer is not allowed to connect the network server, classified servers associated with enterprise audit can't,

CodePudding user response:

This is really bad, LZ have time or training try yourself, stand for the
Font picture is bad hand first, secondly determine what kind of font is not good to train range (such as need be, regular script, etc.), third, training seems to be the biggest type of limited value for the font file (I remember to do, more than 60 kinds of fonts is not supported (guess is training the font too much the last generated font file size is too big, so do the limits, perhaps can consider according to more than 60 kinds of fonts to make the font file after the merger, at that time don't have time to study), so you need what font you want to good), but you require 90% recognition rate, due to the constraints, that is to say, such as identification of font font didn't appear in your training, possible recognition rate of 0,

CodePudding user response:

Use this try?
https://github.com/tesseract-ocr/tesseract/wiki

CodePudding user response:

refer to the second floor KeepSayingNo response:

use this try?
https://github.com/tesseract-ocr/tesseract/wiki

this want to over the wall, access may not