Text to Image Translation using Generative Adversarial Networks

The learning process becomes easier when one can visualize the things being spoken about or being described. To help a person visualize, the description in the form of text which the person gives can be translated to a set of images, this is achieved by a Generative-Adversarial Model. A novel implementation for translating description to images using Generative Adversarial networks is proposed in this paper. We propose a RNN-CNN text encoding along with the Generator and Discriminator network to take the text description of flowers as the input and the resultant output would be a set of unique images generated which match the description for the same. The dataset primarily used is the Oxford 102flowers dataset along with its captions procured from the Oxford University website. It has 102 categories of flowers with each category consisting of a minimum of 40 images.