If I was going to attempt this I would attempt to solve this via pre-processing. Shouldn't be too difficult to pre process to remove the white spaces in the chars to create consistent shapes that tesseract could read easily.
Could possibly need some up-scaling to off set the reduced size too. I don't think this is the answer you are after, but getting tesseract to consider broken shapes as blobs will be tedious. On Friday, March 4, 2016 at 2:23:34 AM UTC+8, Roger wrote: > > Does running tesseract training exhaustive on the .box and .tif files, > helps in the recognition accuracy increase? > > On Wednesday, March 2, 2016 at 4:23:44 AM UTC-3, Roger wrote: >> >> I am training tesseract to recognize CMC7 font, following this >> <http://michaeljaylissner.com/posts/2012/02/11/adding-new-fonts-to-tesseract-3-ocr-engine/> >> and this >> <https://github.com/tesseract-ocr/tesseract/wiki/TrainingTesseract> >> tutorial. >> >> >> I have made a .tif file with 2621 characters, and created the .box file, >> going into every character to make sure the X and Y positions are correct >> (the rectangle around the character). >> >> >> After that, I have run the command to train tesseract: >> >> >> tesseract por.cmc7.exp0.tif por.cmc7.box nobatch box.train .stderr >> >> >> I've made a shell script that calls this command in a loop, so the >> training wil be repeated a bunch of times. However, after a bunch of: >> >> APLY_BOXES: Unlabelled word at :Bounding box=(762,2763)->(783,2776) >> >> APPLY_BOXES: Unlabelled word at :Bounding box=(774,2269)->(783,2277) >> >> APPLY_BOXES: Unlabelled word at :Bounding box=(787,2269)->(789,2277) ... >> >> >> >> The result is always: >> >> Found 420 good blobs. >> >> 2129 remaining unlabelled words deleted. >> >> Generated training data for 420 words >> >> It is running for several hours, and still it generated training data for >> only 420 words. And after I run tesseract on a check image to test it will >> recognize the characters, it doesn't work (doesn't recognize the characters >> and return random letters and symbols). >> >> >> How can I make it recognize all the characters in the .tif image? >> >> >> Thank you. >> >> I have attached the .box and .tif in the zip file. >> > -- 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/1ec2b23b-b424-4adf-a589-3ad3ae0f49f4%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
