Text-To-Image Synthesis Using Modified GANs
Lakshmi S Hanne, R Kundana, R. Thirukkumaran, Yagna Vikas Parvatikar, K Madhura · 2022 International Conference on Advances in Computing, Communication and Applied Informatics (ACCAI) · 2022
Synthesis of high-resolution images for the input textual descriptions is a prominent field of research that needs to be looked into, in this world of computer vision and animations. The task of image generation has its own challenges such as, working on the main object and background elements simultaneously, considering the ambiguity of natural language and reducing the noise that may be collected along with the input. It is also equally important to determine the importance of words given in the input textual description and ranking them according to their degree of relevance to desired image. Generative Adversarial Networks (GANs) are one of the newest developments for synthesis of images from text. In our literature review article, we have provided a consolidated idea of GAN s and how their variants are helpful in producing high resolution images.