A Primer on Conditional Text based Image Generation through Generative Models

Synthesis of Images from text descriptions has emerged as an interesting albeit a challenging task in the domain of Image Synthesis. Many promising advances have been made in the direction of text-based image generation in the recent years, with the emergence of Multi-modal Generative Adversarial Networks. In this paper, we discuss the various approaches which utilise Conditional-GANs to accomplish the task of generating photo-realistic images based on their text descriptions and compare their architectures and performance on various benchmark datasets. The performance of these approaches are evaluated using various well-known metrics.