Recognition Evaluation

[ars22]

Hi,

Thanks for sharing the code. I ran into some NaN issues but was eventually able to train model. When I run eval.py with the trained model it only predicts bbox and not the transcription. In the current code base is it possible to predict the words for each predicted box too? If not what is the easiest way to go about doing it?

Thanks!

[ars22]

ping on this!