An RNN model based on 'Drawing and Recognizing Chinese Characters with Recurrent Neural Network'(Xu-Yao Zhang, Fei Yin, Yan-Ming Zhang, Cheng-Lin Liu, Yoshua Bengio, 2016)
Available at: https://arxiv.org/abs/1606.06539
About this model:
Library used: Keras, numpy
For optional functionality: matplotlib, livelossplot, struct, threading, multiprocessing
Accuracy: 95.38%
Training time: 5 minutes (Intel i5-8259U, 16 GB, macOS Mojave 10.14.4, No dedicated GPU)

IO.py is used to decode .pot files