- python 3.5+
- pytorch 1.1
- torch_baidu_ctc
- lmdn
- opencv
Prepare a text in the following format
/path/to/img/img.jpg helloword
...
- config the
train_file
,test_file
in config.json - generate alphabet
use fellow script to generate
alphabet.py
in the some folder withtrain.py
python3 utils/get_keys.py
- use following script to run
python3 train.py
predict.py is used to inference on single image
- config
model_path
,img_path
in predict.py - use following script to predict
python3 predict.py