Hi Leaf, I suggest you to keep up with the latest versions as much as possible. Not all revisions are stable and can be compiled out of the box. Don't know about the head revision, but for me rev549 (3.01) works just fine. It had a bit of trouble compiling it under MSVC10 but overcoming was relatively easy. The API changes frequently, but imo the functions I've mentioned can hardly be eliminated in the future. Well, at least they are a good point to look how your custom framework should be organized to make Tess think it's being fed with a "bona fide" document.
Warm regards, Dmitri Silaev On Fri, Mar 25, 2011 at 2:38 AM, leaf corcoran <[email protected]> wrote: > Hi, > Thanks for the response. > > I looked for those functions you mentioned and I noticed that they are > only in the SVN. I am working with the tesseract-3.00 source archive > because I was unable to compile tesseract from SVN on OS X. > > But I have made pretty good progress in 3.00. I've been able to get it > to recognize some of my sample glyphs. I've converted them into lists > of line segments which I insert into a PBLOB. From there I create a > WERD and insert it into a ROW, and then a BLOCK. After that I am > running Tesseract::recog_all_words(). I don't have consistent results > yet though, I think this is because I don't have my initial state set > up correctly for the ROW and BLOCK. > > I am wondering now if I should try to get what is on the SVN to > compile. It looks like the API has changed a bit. > > Thanks, Leaf > > > On Mar 24, 2:30 pm, Dmitri Silaev <[email protected]> wrote: >> You can begin your journey to the guts of Tesseract from the >> TessBaseAPI::MakeTBLOB() >> function. Its result then can be used e.g. in LearnBlob() from >> "blobclass.cpp" or >> TessBaseAPI::RunAdaptiveClassifier() >> >> However, if you say you have curves besides line segments, you need to >> restore your glyphs >> in the bitmap form and then run the training/recognition in a usual >> way. This is because afaik Tess >> doesn't support curve approximations of glyph outlines, it only uses >> straight segments. Or, to do >> this programmatically, you might use the >> MakeTBLOB()/RunAdaptiveClassifier() approach. >> >> HTH >> >> Warm regards, >> Dmitri Silaev >> >> >> >> >> >> >> >> On Wed, Mar 23, 2011 at 9:28 PM, leaf corcoran <[email protected]> wrote: >> > I have a collection of unknown glyphs for letters that I am trying to >> > identify in the form of a polygon made of line segments and curves. >> > I'm wondering if tesseract would be a good candidate for this. I've >> > been looking through the source code now but I haven't been able to >> > isolate any part that does what I am trying to achieve. I am using >> > tesseract 3.00 >> >> > I am looking at PBLOB right now, but I haven't seen a way to far to >> > get around the page and word recognition and go right to identifying >> > characters. From what I have read it looks like the classify directory >> > is where I should be looking but I am not so sure. >> >> > Thanks for any help, >> > Leaf >> >> > -- >> > You received this message because you are subscribed to the Google Groups >> > "tesseract-ocr" group. >> > To post to this group, send email to [email protected]. >> > To unsubscribe from this group, send email to >> > [email protected]. >> > For more options, visit this group >> > athttp://groups.google.com/group/tesseract-ocr?hl=en. > > -- > You received this message because you are subscribed to the Google Groups > "tesseract-ocr" group. > To post to this group, send email to [email protected]. > To unsubscribe from this group, send email to > [email protected]. > For more options, visit this group at > http://groups.google.com/group/tesseract-ocr?hl=en. > > -- You received this message because you are subscribed to the Google Groups "tesseract-ocr" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/tesseract-ocr?hl=en.
