基于pytorch卷积神经网络的中文手写汉字识别,使用HWDB数据库
- PIL
- numpy
- torch
- torchvision
- tensorboardX(for visulizztion)
- Download HWDB dataset and unzip to
data
folder - run
python process_gnt.py
to generate img from gnt fiel. Due to the huge dataset (897758+223991 images), it may take a lot of time. I suggest to put the data folder out of project or your pycharm will get slow. - run
python hwdb.py
to visualize the image. - run
python train.py
to start trianing.