The significance of Generative Adversarial Networks (GAN) cannot be overemphasized, especially its adoption in computer vision [1] applications such as image generation, image to image translation, facial attribute manipulation and similar domains. GAN is a generative model in machine learning. The architecture (Figure 1) is made of two networks: Generator and Discriminator [2]. The generator function is basically to create an object that is as close as the real data using a random noise variable as the input. The discriminator on the other hand must be able to differentiate the data coming from the generator and the actual data. The advantages of GAN over other generative models, such as variational autoencoders, is that it can handle sharp estimated density functions, generate desired samples efficiently, and eliminate deterministic bias. However, with the great successes achieved applying GANs to real-world problems possess significant challenges