How to extract text from pictures?

Question

Neural networks

There is a set of various pictures on which there is a text. Pictures are completely different with different backgrounds, etc.
At the output, I want to get the recognized text from the image.
After playing with tesseract, it recognizes pictures that are on a white background relatively well, but disgustingly recognizes unstructured text against the background of some picture.

If I understand correctly, then first you need to use the neuron to find all the letters in the picture - transfer them to a new blank canvas (white background).
And then actually try to recognize it with the help of tesseract.

Tell me, has anyone already tried to solve a similar problem (extracting letters from a picture)?

Answer 1

https://www.pyimagesearch.com/category/optical-cha...