2D Attentional Irregular Scene Text Recognizer

Requirements

• mxnet 1.4
• opencv3
• mxboard

Data Preparation

train: prepare a text in the following format, use '\t' as a separator

/path/to/img.jpg label
...

Train

1. config the train_data_path,val_data_pathin ./configs/eng_config.py or chr_config.py
2. use following script to train

 python3 train.py

Predict

1. modify test_demo.py
2. use following script to predict

 python3 test_demo.py