robbiewu008 / caoyuan-ocr

Optical Character Recognition using Tensorflow and Keras

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:

cd ocr
python3 gen_ocr.py train
python3 gen_ocr.py gen_test_data
python3 gen_ocr.py evaluate

Start server:

FLASK_APP=main.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