deep-text-recognition-benchmark icon indicating copy to clipboard operation
deep-text-recognition-benchmark copied to clipboard

Published 20 hours ago • verified publisher icon clovaai

Suggestions for fine tuning on MRZ ( Machine readable code) images?

Open NavaneethanRajasekaran opened this issue 4 years ago • 0 comments

Awesome work!!!

I can see that current model is trained on Max word length of 25. I am planning to collect some MRZ data which is of 44 characters and train the model. I initially fine tuned the pre trained model with 1k MRZ data and results were not great. I assume that dataset is not big enough. Any suggestions on how i can go head ? Should i retrain only the prediction layer? Please suggest passort example

NavaneethanRajasekaran avatar Jun 02 '21 16:06 NavaneethanRajasekaran