Wrong result for Multi-Line Number Plate in TLT-V3 lprnet

Hi Team,

I am using Nvidia-TLT-V3 and trying to fine-tune lprnet sequence classifier for my custom dataset.
I have trained model successfully but getting wrong result for Multi-Line number plate.
Please find the sample below:

Mline2
Classifier Result : 7U90, Expected Result : MH24AF9090
Mline4
Classifier Result : 3408, Expected Result : WB02AA3408

Below are the images and their labels correspond.
Mline1
Label Assigned during training :
MH26
AK0606
Mline3
Label Assigned during training :
PY05
A5001
Mline5
Label Assigned during training :
TN
23F9023
86
Label Assigned during training :
GJ6
FC7848

Can someone please suggest me why this is happening and also if there is problem with the labeling of the data so please suggest how should I label data for multi-line number plate

.

Currently, LPRnet does not support two lines of text yet.

Thanks @Morganh for the response.

Is there any other LPR model available so that we can fine-tune that for multi line number plate ?

Sorry, there is not. Currently, the lprnet model can only work well on one line of text.