Before doing anything else, load the dataset by performing the following steps:

1. Start by creating a list of image paths, using the glob module as follows:

imagesA = glob(data_dir + '/testA/*.*')
imagesB = glob(data_dir + '/testB/*.*')

We have data from two domains, A and B, which is why we have created two lists.

2. Next, iterate over the lists. Load, resize, and horizontally flip the images inside the loop, as follows:

allImagesA = []
allImagesB = []

# Iterate over the lists
for index, filename in enumerate(imagesA):
     # Load images
     imgA = imread(filename, mode='RGB')
     imgB = imread(imagesB[index], mode='RGB')
    
     # Resize images
     imgA = imresize(imgA, (128, 128))
     imgB = imresize(imgB, (128, 128))
    
     # Randomly horizontally flip images
     if np.random.random() > 0.5:
         imgA = np.fliplr(imgA)
         imgB = np.fliplr(imgB)

     allImagesA.append(imgA)
     allImagesB.append(imgB)

3. Now, normalize the images to bring the pixel values to a range between -1 and 1, as follows:

# Normalize images
allImagesA = np.array(allImagesA) / 127.5 - 1.
allImagesB = np.array(allImagesB) / 127.5 - 1.

The entire code to load the dataset appears as follows:

def load_images(data_dir):
    imagesA = glob(data_dir + '/testA/*.*')
    imagesB = glob(data_dir + '/testB/*.*')

    allImagesA = []
    allImagesB = []

    for index, filename in enumerate(imagesA):
        # Load images
        imgA = imread(filename, mode='RGB')
        imgB = imread(imagesB[index], mode='RGB')
        
        # Resize images
        imgA = imresize(imgA, (128, 128))
        imgB = imresize(imgB, (128, 128))
        
        # Randomly horizontally flip images
        if np.random.random() > 0.5:
            imgA = np.fliplr(imgA)
            imgB = np.fliplr(imgB)

        allImagesA.append(imgA)
        allImagesB.append(imgB)

    # Normalize images
    allImagesA = np.array(allImagesA) / 127.5 - 1.
    allImagesB = np.array(allImagesB) / 127.5 - 1.

    return allImagesA, allImagesB

The preceding function will return two Numpy ndarrays. We will use it to load and preprocess the images before we start the training.