This a funny demo of style transfer of paper ๏ผPerceptual Losses for Real-Time Style Transfer and Super-Resolution
the detail information about this code is in my CSDN blog
python train.py -is_training=True -vgg_model='your vgg model path' -train_data_path='your train dataset' -style_data_path='your style img path' -style_w=100
forexample:
python train.py -is_training=True -vgg_model='vgg16.ckpt' -train_data_path='/train2014' -style_data_path='img/wave.jpg' -style_w=100
you can download the vgg16.ckpt model from the url
note : you can change the degree of style transfer by changing style_w value, and you can also modify the code to set the args papameter as default
python train -is_training=False -test_data_path='your test img' -new_img_name='transfer.jpg' -transfer_model='your saved model after train'
forexample:
python train.py -is_training=False -test_data_path='dog.jpg' -new_img_name='transfer.jpg' -transfer_model='model_saved/wave.ckpt'
if you find it is useful for you, please leave your star, thanks. =_=
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent