Generating Facial Images using VAEGAN

Tento projekt byl podpořen grantem Západočeské univerzity v rámci projektu č. SVK1- 2018-024. Studentská vědecká konference je pořádána s podporou prostředků na specifický vysokoškolský výzkum SVK1-2018-024.A generative modeling is a powerful way how to learn any kind of data distribution using unsupervised learning. The main advantages of generative models are as follows: (1) they generally don’t need any labeling during training; (2) they are able to generate new data similar to existing data. Thanks to these features, generative models become very popular during last three years with a vast field of usage. In this work, I introduce an experiment of generating facial images by using VAEGAN. For the training of my model, I used CasiaWebFace database (Yi et al. (2014)) and CelebFaces Attributes dataset (Liu et al. (2015))