Different Techniques of Facial Image Generation from Textual Input : A Survey

The task of Text-to-Face synthesis is quite intricate and there hasn't been much research done on it, until recently. This is mainly due to the complex nature of human face and it's features, which vary widely over different situations. That being said, the field of Text-to-Image synthesis has gathered considerable interest quite late. In earlier research, generation is mainly done using reconstruction of visuals which correlate to the given words. Due to the rise of generative models in the field of deep learning, there has been a departure from these traditional computer vision based retrieval methods. One of the most significant factors for this change is the introduction of GANs. The idea of learning and reproducing the image as a whole, helped in producing better images. The introduction of attention based mechanisms, helped in synthesizing more detailed images which manages to show several facial features like eye brows, hair color, nose shape etc. In this survey, we discuss and summarize some of the methods used for the purpose of image & face generation and their development over the years.