Is writing code for OCR difficult? It seems like it would be. How much time did it take you to write it?

I was writing this on my spare time so I didn't record how many hours. I would say it took couple of months writing the core algorithm and maybe half a year for the first release. The user interface actually took more time than OCR itself.

vtx4848 Wrote:Is writing code for OCR difficult? It seems like it would be. How much time did it take you to write it?

Anyone use this with OSX? I can get it to run but it seems buggy. If I load image files in a directory for example it will go forward and not backwards and sometimes it doesn't even go forward. Sometimes it just stops working completely. I'm using 2015 macbook pro 16 meg memory with Yosemite 10.10.5 (14F27).

kraemder Wrote:Anyone use this with OSX? I can get it to run but it seems buggy. If I load image files in a directory for example it will go forward and not backwards and sometimes it doesn't even go forward. Sometimes it just stops working completely. I'm using 2015 macbook pro 16 meg memory with Yosemite 10.10.5 (14F27).

I don't think you need to load images with kanji tomo for it to work. You can just load them with any image viewer you want.

Loading the files worked very well on the Windows version for me. I'm thinking of doing Windows Bootcamp on this laptop. Kanjitomo is really nice for studying Japanese and by itself it would make Windows worth it to me. However, it's not just loading the images - I'm having it crash while using it reading PDF's or whatnot on websites too.

Well I have it working well using Simple Comic app to view the manga instead. I've been using it an hour without any problems so I should have tried this from the beginning I think. Anyway - I've found this is a wonderful way to get new vocabulary because I can take partial screen shots and copy and paste those into Anki and then put the vocab from kanji tomo on side2. I just started making a deck with this I'm hoping it's fun to review.

I have released a new version of KanjiTomo. There are no major new features but I have improved the automatic text orientation detection (old version was a bit unreliable) and I have updated the dictionary. There are also some minor bugfixes.

You can download the new version from: http://www.kanjitomo.net

Thanks for the update!

By the way any chance of implementing a wildcard character for searches? Like if you know the second character but don't know the first you have no way of using this program to look it up. It seems like there is always a wildcard at the end by default which is helpful.

Very useful software, thank you.
Ps it's possible to change the dictionary? (Using for example epwing dicts? )
Thanks

Any tips on getting the OCR to work as well as possible? I used to use this program and I remember it being fairly accurate and a real help. I'm in Japan now trying to use it but it's rarely getting the kanji correct now. I'm using it on Apple - maybe that has something to do with it.

(2016-01-26, 2:42 pm)kameden Wrote: By the way any chance of implementing a wildcard character for searches? Like if you know the second character but don't know the first you have no way of using this program to look it up. It seems like there is always a wildcard at the end by default which is helpful.

I have made a small update and uploaded it for download: if you press space in the search text field it finds all words that contain the character. It's not a full wildcard search but maybe enough?

(2016-01-27, 3:59 pm)cescoz Wrote: Very useful software, thank you.
Ps it's possible to change the dictionary? (Using for example epwing dicts? )
Thanks

The dictionary is stored in data/dictionary folder; edict2 is the most important file, it is loaded from http://ftp.monash.edu.au/pub/nihongo/00INDEX.html

If you can generate edict2 file with same format but different content then it is possible to change the dictionary.

(2016-02-07, 6:43 am)kraemder Wrote: Any tips on getting the OCR to work as well as possible?  I used to use this program and I remember it being fairly accurate and a real help.  I'm in Japan now trying to use it but it's rarely getting the kanji correct now.  I'm using it on Apple - maybe that has something to do with it.

KanjiTomo can work differently on Apple because the system fonts are different from ones used in Windows. The font cache is stored in data\cache folder; if you copy these files from a Windows computer to Mac it should work the same way.

Also some source materials are simply easier to identify with OCR. If the text has very small font or non-standard style (especially if it's handwritten) then it will make more mistakes.

(2016-02-10, 12:28 pm)Kurotowa Wrote:

(2016-01-26, 2:42 pm)kameden Wrote: By the way any chance of implementing a wildcard character for searches? Like if you know the second character but don't know the first you have no way of using this program to look it up. It seems like there is always a wildcard at the end by default which is helpful.

I have made a small update and uploaded it for download: if you press space in the search text field it finds all words that contain the character. It's not a full wildcard search but maybe enough?

Thanks, it looks like it works.