CNN-LSTM-CTC text recognition

I realize three different models for text recognition, and all of them consist of CTC loss layer to realize no segmentation for text images.

Disclaimer

I refer to the official mxnet warpctc example here.

Getting started

• Build MXNet with Baidu Warp CTC, and please following this instructions here.

When I use this official instructions to add Baidu Warp CTC to Mxnet, there are some errors because the latest version of Baidu Warp CTC has complicts with mxnet. Recently, I see someone has already solved this problem and updated the official mxnet warpctc example. However, if you still have problem, please refer to this issue here.

Generating data

Run generate_data.py in generate_data. When generating training and test data, please remember to change output path and number in generate_data.py (I will update a more friendly way to generate training and test data when I have free time).

Train the model

I realize three different models for text recognition, you can check them in symbol:

1. LSTM + CTC;
2. Bidirection LSTM + CTC;
3. CNN (a modified model similiar to VGG) + Bidirection LSTM + CTC. Disclaimer: This CNN + LSTM + CTC model is a re-implementation of original CRNN which is based on torch. The official repository is available here. The arxiv paper is available here.

• Start training:

LSTM + CTC:

python train_lstm.py

Bidirection LSTM + CTC:

python train_bi_lstm.py

CNN + Bidirection LSTM + CTC:

python train_crnn.py

Prediction

You can do the prediction using your trained model. I only write the predictors for model 1 and model 3, but it is very easy to write the predictor for model 2 when referring to the examples.

Plesae run:

python lstm_predictor.py

or

python crnn_predictor.py