Residential layout generation based on land boundary conditions

Abstract

Residential groups are the work with strong regularity and large calculation, and often requires a lot of time to write procedures. The method of deep learning in the field of artificial intelligence can learn the rules in the data based on the training of a large amount of data, and generate new data with the same rules to save labor costs and improve design efficiency. This article studies how to generate corresponding residential groups based on the specified land border.

The research is mainly based on the Pix2pix network in deep learning. Pix2pix can generate the corresponding image according to the given condition image. Using Pix2pix to learn the schemes of a large number of residential groups, the network after learning can quickly automatically generate the residential group scheme diagram according to the given condition image, thereby assisting the design.

The research is mainly divided into three steps. The first step is to extract the land boundary and building outline images for training based on OpenCV; the second step is to conduct model training based on pix2pix to obtain the residential layout plan generated based on the land boundary. The third step is to use grasshopper to visualize the results and read the coordinates of the contour points to generate a model of the solution.

The main purpose of the research is to use deep learning methods to assist architects in residential layout design and improve design efficiency.