This is a tensorflow re-implementation for the paper: "EAST: An Efficient and Accurate Scene Text Detector"
More details: https://arxiv.org/pdf/1704.03155.pdf
- Python3
- tensorflow
- numpy
- opencv-python
- shapely
Dependencies can be installed with
pip install -r requirements.txt
- Put all images in ./data/img/
- Put corresponding ground truth in ./data/gt
For example:
./data/img/1.jpg
./data/gt/1.txt
data format: Link
python data_processor.py
python train.py
python valid.py