def _save_latent_space(self) -> None:
latent_space_inputs = np.zeros(
(self._latent_space_rows * self._latent_space_columns,
self._latent_dim))
for i, v_i in enumerate(
np.linspace(-1.5, 1.5, self._latent_space_rows, True)):
for j, v_j in enumerate(
np.linspace(-1.5, 1.5, self._latent_space_columns, True)):
latent_space_inputs[i * self._latent_space_rows +
j, :2] = [v_i, v_j]
generated_data = self._generator.predict(latent_space_inputs)
plot_save_latent_space(generated_data, self._latent_space_rows,
self._latent_space_columns, self._resolution,
self._channels, self._outputs_dir, self._epoch)
def _save_latent_space(self) -> None:
latent_space_inputs = np.zeros(
(self._latent_space_rows * self._latent_space_columns,
self._latent_dim))
random_class = np.random.randint(0, self._classes_n)
latent_space_classes = np.ones(
self._latent_space_rows *
self._latent_space_columns) * random_class
latent_space_classes = to_categorical(latent_space_classes,
self._classes_n)
for i, v_i in enumerate(
np.linspace(-1.5, 1.5, self._latent_space_rows, True)):
for j, v_j in enumerate(
np.linspace(-1.5, 1.5, self._latent_space_columns, True)):
latent_space_inputs[i * self._latent_space_rows +
j, :2] = [v_i, v_j]
generated_data = self._generator.predict(
[latent_space_inputs, latent_space_classes])
plot_save_latent_space(generated_data, self._latent_space_rows,
self._latent_space_columns, self._resolution,
self._channels, self._outputs_dir, self._epoch,
self._classes[random_class])