Geração de modelos de elevação digital a partir de esboços topográficos utilizando redes generativas adversárias

Abstract

Although extensively exploited, creating terrain in real time is still a challenge. The advancement of computational power through graphics processing units arose the artificial neural networks. This work presents a new method for synthesizing terrain interactively using generative adversary networks. Digital Elevation Models of the world and their respective topographies are used for the synthesizer network’s training. The proposed solution enables the real-time generation of realistic high-resolution terrains that follow the nature of the topographical sketches inputs.