Scene generation from backgrounds to objects and anything in between: A deep learning robotics survey
The recent rapid progress of deep learning algorithms in generating realistic images, especially in Generative Adversarial Networks (GAN) and Variational Auto-Encoders (VAE), has helped advance new applications. Examples of such applications range from generating and manipulating new synthetic data for self-driving cars, to building/urban architectures, to interior design, and gaming. Furthermore, several applications have benefited from deep learning generative advancement, such as robotics manipulations in structured and unstructured environments, virtual fashion clothes try-on, and item identification on the go. This survey paper provides a review of techniques for image generation from background outdoor scenes, to building facades and objects, and anything in between. In particular, we will cover scene generation such as outdoor landscapes, building facades and indoor scenes. For each category, we will compare the existing state of the art algorithms and techniques, and discuss their performance and gaps limitations on a wide variety of inputs. Additionally, we will discuss challenges and future trends to advance the state of the art in realistic image generation.