SRL Archive - Research & Development Lounge

TesseractTool - A Helper tool for finding the best (simple) filter to get the best recognition results.

I wrote a small tool in Lape so people can understand, and figure out how the simple filters we apply to images for Tesseract to work optimal.

Optimal way is to re-size > threshold > Tesseract.

Couple of tips:

• Left click on image to start to draw a box, left click again to finish the box.
• Press enter to crop the image.
• Control + v to paste from clipboard.
• Set whitelist (Chars to only read) under the tesseract menu.

To-do:

• Undo/Redo

Git: https://github.com/Olly-/TesseractTool


Also, Thanks to slacky, He made me release it.. and helped with bit of it.

Thanks for this, especially since I didn't know exactly what the constants were doing, I had been guessing and checking up until now.

Nice! This can come in handy!

Great work Olly :)

Didn't you release this ages back? After moar reps i c :p

Great tool!

Would this tool help with text recognition of AA text? If not what is this used for in text recognition? Seems interesting :).

From: elephant
Would this tool help with text recognition of AA text? If not what is this used for in text recognition? Seems interesting :).

Yes, probably. This uses TesseractOCR (https://code.google.com/p/tesseract-ocr/) and Tesseract is also used in SRL-6 on somethings where handmade fonts (as well as simba's method) just dont cut it.

There is nothing really required to do, ie font picking, sizing, spacing, etc, just give tesseract a bitmap (or in SRL-6's case, a tbox) and it extracts the text fairly well.

Keep in mind, as this is not like simba's OCR, it is slower. But for AA'ed text, it will probably work better.

Would this tool help with text recognition of AA text? If not what is this used for in text recognition