Which recognition method is relevant for recognizing handwritten letters (Cyrillic)?
If the task is to recognize handwritten letters, which method/algorithm of machine learning is most relevant and gives more accurate results in this area at the moment? For example, is it better to use neural networks, or can machine learning methods be completely dispensed with without INS?
Now I am faced with the task of recognizing handwritten letters (Cyrillic). The letters will be recognized individually (not the whole words). I will submit tables for input with various handwritten letters of the Russian alphabet. We need help in choosing the most optimal algorithm/method for recognizing handwritten characters.
2 answers
When it comes to image recognition, the answer immediately arises - a convolutional neural network. Your case is no exception. I must say that there are many ready-made solutions with a ready-made NS architecture, so you do not need to write anything yourself. If C++ or Python, I recommend Tensorflow. To understand how Convolutional NS works, read my answer to a similar question.
You don't need any neural networks, especially convolutional ones.
Take Tesseract, train on handwriting and use it.
Tesseract itself uses NS, but not convolutional :-)
Most recently, I made a project-Tesseract is used to detect charboxes, then Tesseract tries to recognize what it has detected, in parallel, the same charboxes tries to recognize an algorithm based on OpenCV kNN.
For what is not recognized, there is a GUI in which you can select the charbox and manually specify what character it is. As a result, data is generated for the tesseract training (what it could not recognize itself and was entered manually), and data for the kNN training (what tesseract recognized and what was manually entered). After >10 training sessions per symbol, it starts to be recognized fairly steadily by both systems.