In [4]:

    

import numpy as np
import tensorflow as tf
import pandas as pd
from scipy.io import loadmat


    

In [5]:

    

###############################################################################
# Constants for the image input and output.
###############################################################################

# Output folder for the images.
OUTPUT_DIR = 'output/'
# Style image to use.
STYLE_IMAGE = 'styles/shipwreck.jpg'
# Content image to use.
CONTENT_IMAGE = 'input_images/lion.jpg'
# Image dimensions constants. 
IMAGE_WIDTH = 800
IMAGE_HEIGHT = 600
COLOR_CHANNELS = 3


    

In [6]:

    

###############################################################################
# Algorithm constants
###############################################################################
# Noise ratio. Percentage of weight of the noise for intermixing with the
# content image.
NOISE_RATIO = 0.6
# Constant to put more emphasis on content loss.
BETA = 5
# Constant to put more emphasis on style loss.
ALPHA = 100
# Path to the deep learning model. This is more than 500MB so will not be
# included in the repository, but available to download at the model Zoo:
# Link: https://github.com/BVLC/caffe/wiki/Model-Zoo
#
# Pick the VGG 19-layer model by from the paper "Very Deep Convolutional 
# Networks for Large-Scale Image Recognition".
# http://www.vlfeat.org/matconvnet/pretrained/
VGG_MODEL = './neural_style_painting/imagenet-vgg-verydeep-19.mat'
# The mean to subtract from the input to the VGG model. This is the mean that
# when the VGG was used to train. Minor changes to this will make a lot of
# difference to the performance of model.
MEAN_VALUES = np.array([123.68, 116.779, 103.939]).reshape((1,1,1,3))


    

In [ ]: