MINGHAO XU 

Generated by Prediction

An Urban Design Technology Assisted by Big Data and Generative Adversarial Network

Undergraduate Program @ Southeast University

Team: Minghao Xu, Yifan Cui, Jian Wen.

Instructor: Li Li

Employed big data and Generative Adversarial Network (GAN) in the urban design of a tourist town near Nanjing, leveraging extensive city data and precedents to address development intensity, building type, and architectural form, leading to streamlined design cycles and enhanced economic precision in designs.

​

China's urban development has left the stage of rough expansion, which puts forward higher requirements for urban design. Slogan-style and artistic conception-style design methods can no longer meet the needs of developers. Therefore, a quantitative urban design technique is urgently needed to guide designers.

With the development of a tourist town around Nanjing as the background, I used big data and GAN to help the target town solve some quantitative problems in urban design. By researching a large amount of city data and many mature and similar precedents, it can provide references for new design.

This technology can help solve problems including development intensity, building type and architectural form in the early stage of land development, reducing the design cycle as much as possible and improving the economic accuracy of the design.

Background and method

Sample selection and statistical prediction

Pix2pix prediction of POI distribution and urban texture

Use of predictions to generate possible urban massings