Testing the models

To test the networks, create the generator and the discriminator networks. Then, load the learned weights. Finally, use the predict() method to generate predictions:

# Create models
generator = build_generator()
discriminator = build_discriminator()

# Load model weights
generator.load_weights(os.path.join(generated_volumes_dir, "generator_weights.h5"), True)
discriminator.load_weights(os.path.join(generated_volumes_dir, "discriminator_weights.h5"), True)

# Generate 3D images
z_sample = np.random.normal(0, 0.33, size=[batch_size, 1, 1, 1, z_size]).astype(np.float32)
generated_volumes = generator.predict(z_sample, verbose=3)

In this section, we have successfully trained the generator and the discriminator of the 3D-GAN. In the next section, we will explore hyperparameter tuning and various hyperparameter optimization options.