Nativescript core Firebase ML kit offline text recognition vertical

Question

Are there any method for Nativescript core with offline Firebase ML kit for vertical text recognition? It works really good for horizontal text, but I need to scan maritime containers and some of them have the id in vertical mode. It seems to have problems with independant letters.

Example: (I need to get the text of the right)

Another example:

I have tried it with an old Huawei and Samsung Galaxy s8 so I think that is not a camera problem.

It seems not to be problem of the picture, because I try with a simple text on a .txt like this:

A
B
C
U
1
2
3
4
5
6
0

and it doesn´t works it only reads like:

ABCU1234560

There must be a single characters problem because if I edit the first picture converting Id in horizontal format it works, like in the example:

But I need to catch the vertical one in offline mode

I have just tried in a microsoft word doc and it works fine with this:

AA
BB
CC
UU

and it doesn't works for

A
B
C
U

I looked for any kind of param or config in the plugin files who limits the chars to "more than one", but it semms not to be anything with that funcionality. (sometimes it can scan for one letter like "A", but not the other ones except if there are more than one char.

Now I tried with this with good results:

AA A
BB  B
CC   C
UU    U

It seems that the ocr doesn't detect when it have to start.

Any idea about how to force a start recognicing when there are only one char?

Answer 1

Only can be done with online mode (that works perfect) for this case. Could solve this saving the photos in local gallery and recover them when can access to internet to do the stuff. Is not clean, but is the unique method without online optimized recognition.