Real results

[deflorator1980]

In my occasion real results are pretty sad :)

[Wikunia]

I suppose that the sliding windows you are using are too big. You can either change that or increase the size of the photo.
In general there are better approaches than sliding window. If it's just black on white you can use bounding boxes generated by openCV. You might wanna check my He2LaTeX repository for that.

[Wikunia]

Any news?

[deflorator1980]

Any news?

He2LaTeX (https://github.com/Wikunia/HE2LaTeX) detects my handwritten digits even worse )

[Wikunia]

It is trained using different equations and I found out that it isn't generalizing it. When running it in verbose mode you should get a good output of the single digits ;)
Sorry for the results :/
With my comment I meant more that the bounding boxes should be better in that code.
I should probably add it to this code and if I have time relearn on He2LaTeX

[deflorator1980]

It is trained using different equations and I found out that it isn't generalizing it. When running it in verbose mode you should get a good output of the single digits ;)
Sorry for the results :/
With my comment I meant more that the bounding boxes should be better in that code.
I should probably add it to this code and if I have time relearn on He2LaTeX

I my case YandexVision works pretty good https://cloud.yandex.com/docs/vision/quickstart
I don't know what library do they use under hood:)