Training OCRNet for being used for LPD/LPR

For @foreverneilyoung 's shared license plate,
image
I confirm that OCRNet_vit model can recognize the “B JT 304 E” including the spaces.
But need to retrain a new model with this kind of datasets(including spaces).