load-data

# Loads the images for use with the CNN.
def load_images(image_size=32, batch_size=64, root="../images"):
    transform = transforms.Compose([
        transforms.Resize(32),
        transforms.ToTensor(),
        transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])

    train_set = datasets.ImageFolder(root=root, train=True, transform=transform)
    train_loader = torch.utils.data.DataLoader(train_set, batch_size=batch_size, shuffle=True, num_workers=2)

    return train_loader


# Defining variables for use with the CNN.
classes = ('daisy', 'dandelion', 'rose', 'sunflower', 'tulip')
train_loader_data = load_images()

# Training samples.
n_training_samples = 3394
train_sampler = SubsetRandomSampler(np.arange(n_training_samples, dtype=np.int64))

# Validation samples.
n_val_samples = 424
val_sampler = SubsetRandomSampler(np.arange(n_training_samples, n_training_samples + n_val_samples, dtype=np.int64))

# Test samples.
n_test_samples = 424
test_sampler = SubsetRandomSampler(np.arange(n_test_samples, dtype=np.int64))