Generative modeling techniques such as Variational Autoencoder(VAE) and Generative Adversarial Network(GAN), which are rapidly developing in the field of deep learning, have been applied to topology optimization. VAE is a generative modeling technology that extends Autoencoder and is a methodology for generating new images through limited latent space. We modified the general VAE structure to be able to learn the results of the topology optimization design. GAN is another generative modeling technique for generating images, implemented by a system of two neural networks contesting with each other to make realistic synthetic data. It is possible to apply GAN to obtain a more detailed optimization result from the original. The machine learning methodology using the two generative modeling techniques of deep learning is expected to increase the efficiency of topology optimization significantly.