Text recognition with Pytorch Using CRNN and CRAFT pretrained models.
Main sources:
- Character Region Awareness for Text Detection Paper
- [CRNN Paper]
Install it, install requirements.txt by
python -m pip install requirements.txt
and check argparsers. at the moment it is only input image so
> source venv/Scripts/activate
> python app.py --input_file test_image.jpeg
Example results
return .json file with
{"0": "dollar", "1": "glen", "2": "and", "3": "campbeli", "4": "castle"}�0
and wil save it in default path.
python setup.py install
to do :
add docker/flask/redis to store and visualize results
Flask app Bootstrap frontend
docker compose
onxx it and put on mobile. how to make it faster? (todo)
0.1a
Input image (TODO: Preprocess origin format) Output - file with predicted words on input file
Image -> CRAFT -> CRNN -> .json
added simple flask app.