How to implement a neural network for character recognition?

Question

Pattern recognition

Started studying artificial neural networks. I want to implement a program for recognizing black characters on a white background. I chose Typescript as the language.
What can you advise a beginner in this field, what tips for practical implementation can you give?
I found an article on Habré, which says that you need to go through the pixels. Therefore, the question is interesting, how to read pixels using Javascript?
PS I watched a video course on neural networks, but nothing is said about the practical implementation, that is, how it should look in the code.

Answer 1

To begin with, the easiest way to implement something is to understand the finished library. For example ConvNetJS .
Here is an example of MNIST character recognition in JS.

Answer 2

in general canvas
html2canvas.hertzen.com/documentation.html#how-it-works
but this is hardly a good use for js

Answer 3

To begin with, I advise you to switch to C ++. There are good libraries for cross-platform, for example Qt, why limit yourself to scripting capabilities?

Answer 4

Use python or R, lots of examples and ready-made libs.