CRNN

original paper arxiv. The CRNN model can be seen as following:

File	Description
Model .py	Network using CNN (VGG) + Bidirectional LSTM
Model_GRU. py	Network using CNN (VGG) + Bidirectional GRU
Image_Generator. py	Image batch generator for training
parameter. py	Parameters used in CRNN
training. py	CRNN training
Prediction. py	CRNN prediction

Training

• First download the korean license plate data from here

• Put the images in DB/train like following:

  DB/
    train/
      99ak0449.jpg
      99wj0971.jpg
      ...
  

  The filename(99wj0971) is the label. The image can be seen by:

  

• Refer to parameters.py and begin training by python training.py.

Results

• See Prediction.py for prediction on images.

• Some result can be seen by: