How to create a specialized OCR solution for machine vision?

Question

We need to read some text from sales photos received by the iPad camera. Here is an example similar to what we need to read:

There are several limitations to this problem:

• We need to read the total amount that always appears after the text marker (for example, Grand Total in this example).
• The font is always the same.
• The application should work autonomously without a network connection.

This is what we have tried so far:

• Extracting text Google Mobile Vision worked like magic. But text extraction is only available on Android. And we need to build a solution in iOS.
• Google Microsoft , . .
• Tesseract OCR. . , - .

NN. , , , , ?

• . .
• .

, .

• .
• conv net ( ) . .

, . , .

PS. , , . , SO.

Answer 1

deeplearning4j.org. , android. , Android java.