/caoyuan-ocr

Optical Character Recognition using Tensorflow and Keras

Primary LanguagePython

Optical Character Recognition using Deep Learning technologies

Development environment (see Instructions)

  • Ubuntu 16.04 64bit
  • Python 3.5
  • CUDA 8.0
  • CUDNN 6.0
  • OpenCV 3.3
  • Tensorflow 1.3
  • Keras 2.0
  • MongoDB 3.5

Install dependencies:

git clone https://github.com/sunjl/caoyuan-ocr.git
cd caoyuan-ocr
sudo pip3 install -r requirements.txt

Execute tasks:

python3 font_ocr.py train
python3 font_ocr.py gen_test_data
python3 font_ocr.py evaluate

Start server:

FLASK_APP=app.py flask run

Todo:

  • Algorithms:
    • Object Detection: Faster-RCNN, R-FCN, SSD, CTPN
    • Semantic Segmentation: FCN
    • Text Recognition: LSTM, CTC, Attention
  • Support card/license/identification/receipt/table