Optimal Design Scheme of Facade Pane Framing Based on GAN Algorithms
Liu Hangyu, Liang Xinyue, You Jiazhi, Zhang Jinhua,
Lu Baiheng, Shen Jingjie, Bai Bei.
Zhong Ximing, Li Kun
This project is an optimization scheme for dividing building window panes based on CNN and GAN algorithms.
First, collect 40 pictures of scenery outside the window with the same specification parameters as input samples, artificially separate the 40 photo samples, and collect 40 separated photos as output samples.
Secondly, 40 samples are imported into the machine based on GAN for training, and after the training is completed, other photos are input for testing.
Finally, the image data is imported into the design model to design a complete pane. After completing the whole process, we package the program for actual auxiliary design.
The results of machine learning will be used in the design of an actual building window.
This research project is a series of outcomes from the AI architecture course launched by Archiford and Xi’an University of Architecture and Technology. The course aims to use deep learning methods to design and advance architectural design solutions.
GAN is the most widely used technique in the 2D image generation part. It is a machine learning algorithm that trains two neuronal networks to compete with each other in a zero-sum game, as well as developed by later researchers in various directions.
In this project, 40 window view images with the same specification parameters are collected as input samples, and the images of the façade are deconstructed and modal copies are made by GAN to fit the form of the window frame to the view outside the window, constituting the most suitable separation for the view.