Multimodal-source image generation with deep learning

The focus of this work almost completely carried on during my stay at KU Leuven University as part of my Erasmus project, is to present a new way of synthesizing images starting from a descriptive input text and a reference image by using a model, RaGAN, extensively based on deep learning and generative adversarial network, that could set a base for future experimentation in this hybrid field that is multimodal source image generation.