DeepScope: HCI Platform for Generative Cityscape Visualization
Ariel Noyman, Kent Larson · 2020
Streetscape visualizations are necessary for the understanding and evaluation of urban design alternatives. Alongside blueprints and textual descriptions, these design aids can affect city-form, building-codes and regulations for decades to come. Yet despite major advancements in computer graphics, crafting high-quality streetscape visualizations is still a complex, lengthy and costly task, especially for real-time, multiparty design sessions. Here we present DeepScope, a generative, lightweight and real-time HCI platform for urban planning and cityscape visualization. DeepScope is composed of a Generative Neural Network (DCGAN) and a Tangible User Interface (TUI) designed for multi-participants urban design sessions and real-time feedback. In this paper we explore the design, development and deployment of the DeepScope platform, as well as discuss the potential implementation of DeepScope in urban design processes.