Draft version. Production version will be at https://github.com/IlyaOvodov/AngelinaReader (under construction)
ubuntu, windows with GPU
CUDA
Python 3.6
PyTorch 1.4
torchvision
ignite
numpy
PIL
albumentations
cv2
https://github.com/IlyaOvodov/pytorch-retinanet
https://github.com/IlyaOvodov/OvoTools
https://github.com/IlyaOvodov/labelme (for annotation)
git clone https://github.com/IlyaOvodov/BrailleV0.git
edit BrailleV0/local_config.py to set data_path and global_3rd_party pointing to current dir
git clone https://github.com/IlyaOvodov/OvoTools.git
cd OvoTools
python setup.py develop
cd ..
git clone https://github.com/IlyaOvodov/pytorch-retinanet.git pytorch_retinanet
download model https://yadi.sk/d/GW0qEmA5rL0m0A into ./NN_saved/retina_chars_eced60/models/clr.008
cd BrailleV0/web-app
python angelina_reader_app.py
access it by 127.0.0.1:5000
or edit if __name__=="__main__" section BrailleV0/NN/RetinaNet/infer_retinanet.py. Set img_filename_mask and results_dir to proper values and run it.