Learning to Generate Images of Outdoor Scenes from Attributes and Semantic Layouts

• DOI: 10.48550/arxiv.1612.00215
• Publication Date: 2016-01-01
• Authors: Levent Karacan; Zeynep Akata; Aykut Erdem; Erkut Erdem
• Keywords: Minimum bounding box; Bounding overwatch; Generative model

Automatic image synthesis research has been rapidly growing with deep networks getting more and expressive. In the last couple of years, we have observed images digits, indoor scenes, birds, chairs, etc. being automatically generated. The expressive power generators also enhanced by introducing several forms conditioning variables such as object names, sentences, bounding box key-point locations. this work, propose a novel conditional generative adversarial network architecture that takes its strength from semantic layout scene attributes integrated variables. We show our is able to generate realistic outdoor under different conditions, e.g. day-night, sunny-foggy, clear boundaries.