These codes is used by myself and not well organized yet. :-p

1. prepare transcriptions
   • utils/xxx_trans.py
2. prepare allwords.txt in dataset dir
3. generate files used in lang
   • utils/gen_lang_files.py
4. set path in config.py
5. run pipeline.py to generate data
6. run kaldi/run.sh
   • !Remember to set hmm_root and dataset
7. run utils/visaligns.sh
   • show alignments
   • crop chars
8. generate sample list for caffe with utils/gen_list.sh

1. run caffe/makeds.sh
2. train digit model
   • cd caffe
   • ./train.sh

1. run rnnlib/create_nc.py in the folder with config.py