Preprocess the Images

Each image in the edges2handbags dataset folder contains the picture of the bag and the picture of the bag edges in the same image. For training the network we need to separate them as images belonging to the two domains A and B that we have discussed in the architecture for DiscoGAN. The images can be split up into Domain A and Domain B images by using the following code (image_split.py):

# -*- coding: utf-8 -*-
"""
Created on Fri Apr 13 00:10:12 2018

@author: santanu
"""

import numpy as np
import os
from scipy.misc import imread
from scipy.misc import imsave
import fire
from elapsedtimer import ElapsedTimer
from pathlib import Path
import shutil 
'''
Process the images in Domain A and Domain and resize appropriately
Inputs contain the Domain A and Domain B image in the same image
This program will break them up and store them...