Text recognition with Convolutional Recurrent Neural Network and TensorFlow 2.0 (tf2-crnn)

Implementation of a Convolutional Recurrent Neural Network (CRNN) for image-based sequence recognition tasks, such as scene text recognition and OCR.

This implementation is based on Tensorflow 2.0 and uses tf.keras and tf.data modules to build the model and to handle input data.

To access the previous version implementing Shi et al. paper, go to the v.0.5.2 tag.

Installation

tf_crnn makes use of tensorflow-gpu package (so CUDA and cuDNN are needed).

You can install it using the environment.yml file provided and use it within an environment.

conda env create -f environment.yml

See also the docs for more information.

Try it

Train a model with IAM dataset.

Create an account

Create an account on the official IAM dataset page in order to access the data. Export your credentials as enviornment variables, they will be used by the download script.

export IAM_USER=<your-username>
export IAM_PWD=<your-password>

Generate the data in the correct format

cd hlp
python prepare_iam.py --download_dir ../data/iam --generated_data_dir ../data/iam/generated
cd ..

Train the model

python training.py with config.json

More details in the documentation.