In this project, generation of synthetic images is done using Generative Adversarial Networks (GANs) on top of Deep Convolutional layers, i.e., Deep Convolutional GAN[DCGAN]
Sampling from a complex, high-dimensional training distribution of the Fashion MNIST images. Sampling data from gaussian function is done, such as Gaussian noise(since direct sampling from high-dimensional training distribution is inefficient and complex). The model used power of neural networks to learn a transformation from the simple distribution(gaussian noise) directly to the training distribution. The GAN consists of two adversarial players: a discriminator and a generator. The two players jointly play a minimax game to give output.