Face Completion Using Generative Adversarial Network

In this paper, we present a novel idea of hybrid algorithm a two-stage generative adversarial networks (GANs) each comprising of convolutional neural network (CNN) tackling the problem of face inpainting. In order to address various peculiarities of face, we aid the user by accepting textual inputs for the same. We intend to achieve the results by applying pixel generation using GAN (Stage 1) combined with text-to-image conversion using GAN (Stage 2) benchmarking on a custom-made dataset for this task. The details of the implementation and the intermediate experiments along with the recorded observation, data curation and data preprocessing step are also presented.