Hands-On Generative Adversarial Networks with Keras

Interpolation in the Latent Space

We continue our experiments by performing linear interpolation in latent space. We do this by sampling two Z vectors, which are represented as points in a multidimensional space. We interpolate between then by linearly going from one point to another in steps. Let's take a look at the code for doing so:

def infer(data_filepath='data/flowers.hdf5',           z_dim=128, out_dir='gan', n_steps=10):    # load the saved model    G = load_model(out_dir)    # get text embeddings and text from the validation set    val_data = get_data(data_filepath, 'train')    val_data = next(iterate_minibatches(val_data, 1))    emb_fixed, txt_fixed = val_data[1], val_data[2]    # sample two z vectors to mark our starting and ending z z_start = np.random.uniform(-1, ...