Добрый день!
Совершенно неожиданно выяснилась проблема -- оказывается для участия в GSoC иногда вправду сложно найти mentor'а.
Имею предложение в GNU Project следующего содержания:
======
Hello!
I have been try to write own OCR System already for 4 years.
I have been try many approaches to recognition and now I want to use the next idea for it:
I think to use potrace to convert raster image to vector and then handle vector, not raster.
The general problem to recognize cursive text is the splitting of symbols one from ones.
My idea is that we don't need to do it. We can to handle the joined characters. And then we need to recognize split high and low parts of symbols (high part is part which upper than joint line, low part is part which lower joint line).
Now I have taken Title of my qualification project in my university: "The cross-platform handwritten Optical Character Recognition System".
I have written short code in Prolog to try how my idea will work. I hard-code two russian letters and try to find it on the image. The results was wholly satisfactory for me.
But now I study Lisp and I want to use Lisp for my future project.
In bounds of my qualification project I think to have OCR tool for russian texts with hard-coded symbols.
But in GSoC bounds I think to write the editor for symbols and then they will not hard-coded -- they will easy edited.
It will comfortable that we will not need to train the system, but we will only edit it's "fonts" for recognition new symbols.
Also there will feature to recognize math formulas and many comfortable features to easy document input.
I want, the project would be part of The GNU Project. So I select The GNU Project as mentoring organization.
I know, the project idea is insane, so I have another proposal: "Parted. Universal repartition tool for parted."
I hope that at least one of these project will interesting for GNU-society and then I will glad to work for it.
I think that before 1 april I present for you my first simple Lisp code to demonstrate future features of my system. So you would decide what from my two projects will better.
[PLAN]
So, in bounds of my qualification project in university I must to realize OCR tool for russian texts with hard-coded russian symbols. I will write my qualification project until end of May.
At the end of May I continue work for the project in GSoC bounds.
At mid-term I plan to have the OCR for russian-english texts with symbols coded through special "fonts"-files.
So as I will handle image as vector there are will be used paths splitting for separating image over words.
Only after that I can merge my code to the Ocrad. I think it's will good idea to use Ocrad's algorithms of layout analyser, image/frame detections..
So at the end of August I assume to have improved Ocrad which may use two algorithms for character recognizing -- old and new.
The new algorithm will used when old algorithm has bad results.
Advantages of new algorithm:
- It doesn't sense to rotating, scaling of symbols and many other non-linear typical for cursive transformations.
- It is easy to add new symbols for recognizing.
Disadvantages of new algorithm:
- It sense to symbols tears or extra links in symbols. Maybe later there will be algorithms to avoid this problems.
Thanks for attention.
Sorry for bad english
======
Связываюсь с разработчиком Ocrad -- единственного кто мог бы реально быть там наставником и узнаю, что он просто не имеет времени для этого (чтобы быть наставником надо быть доступным несколько чаще чем, просто переписываться по почте раз в сутки). Кроме того он сетует на своё незнание Lisp'а.
Итого получаю ответ:
"Hi Nikolaj Krivchenkov -- mentors have to have more time than Antonio does. So that won't work out. I don't know of anyone else in GNU could mentor you with an OCR project. So I can only encourage you to look at the other SoC organizations."
Пытаюсь найти другую организация.. Взгляд мой падает на LispNYC.. Выбор логичный так как всё же на Lisp пишем..
Только что в чате имею диалог:
- I don't know who could mentor it. submit an application, and if we can find a mentor, it should have a good chance of being accepted
- And if you can't?
- well, then we can't :) but at least you know you tried
В результате.. такая вот дилема..
Понимаю что мой английский плох, но оставляю основной текст на английском, так как меня интересует ответ людей всё же знающих английский, которые реально могли бы быть mentor'ом GSoC.
>>> Подробности (Invalid URL, no host part!)