implementation of neural style transfer in Tensorflow
Input image:
Style image:
Output image:
Parameters:
-
weights of all layers for style loss= 0.2 as per paper
-
alpha=200
-
beta=5
-
optimizer=Adam:epsilon=6
-
based on the VGG 16 network - you can download the weights here: https://drive.google.com/open?id=0B3s6MjhdBGh_VFZtVDFSSzNDYmc
References: