Conditional-Instance-Norm-for-n-Style-Transfer

Implementation of the paper A Learned Representation for Artistic Style

Introduction

Simply implementing the paper A Learned Representation for Artistic Style (Conditional instance normalization)

def conditional_instance_norm(x, scope_bn, y1=None, y2=None, alpha=1):
    mean, var = tf.nn.moments(x, axes=[1, 2], keep_dims=True)
    if y1==None:
        beta = tf.get_variable(name=scope_bn + 'beta', shape=[x.shape[-1]], initializer=tf.constant_initializer([0.]), trainable=True)  
        gamma = tf.get_variable(name=scope_bn + 'gamma', shape=[x.shape[-1]], initializer=tf.constant_initializer([1.]), trainable=True) 
    else:
        beta = tf.get_variable(name=scope_bn+'beta', shape=[y1.shape[-1], x.shape[-1]], initializer=tf.constant_initializer([0.]), trainable=True) # label_nums x C
        gamma = tf.get_variable(name=scope_bn+'gamma', shape=[y1.shape[-1], x.shape[-1]], initializer=tf.constant_initializer([1.]), trainable=True) # label_nums x C
        beta1 = tf.matmul(y1, beta)
        gamma1 = tf.matmul(y1, gamma)
        beta2 = tf.matmul(y2, beta)
        gamma2 = tf.matmul(y2, gamma)
        beta = alpha * beta1 + (1. - alpha) * beta2
        gamma = alpha * gamma1 + (1. - alpha) * gamma2
    x = tf.nn.batch_normalization(x, mean, var, beta, gamma, 1e-10)
    return x

How to use

1. Download the dataset MSCOCO, and unzip the dataset to the folder 'MSCOCO'

├── imgs
├── results
├── save_imgs
├── save_para
├── style_imgs
├── vgg_para
├── MSCOCO
     ├── COCO_train2014_000000000009.jpg
     ├── COCO_train2014_000000000025.jpg
     ├── COCO_train2014_000000000030.jpg
     ├── COCO_train2014_000000000034.jpg
     ├── COCO_train2014_000000000036.jpg
     ├── COCO_train2014_000000000049.jpg
     ...

2. Download the vgg16.npy, and put it into the folder 'vgg_para'
3. Execute the python file 'main.py'

Requirement

• python3.5
• tensorflow1.4.0
• scipy
• numpy
• pillow

Results

Style = alpha * style2 + (1 - alpha) * style1

Content	Style1	Style2	Result

Content	Style1	Style2	Result

Content	Style1	Style2

alpha=0	alpha=0.6	alpha=1.0

Content	Style1	Style2

alpha=0	alpha=0.6	alpha=1.0

Content	Style1	Style2

alpha=0	alpha=0.2	alpha=0.4	alpha=0.6	alpha=0.8	alpha=1.0

Content	style	result