Hybrid Graph-based Generative Architecture of Schematic Floor Plans

Our research provides time-cost and financial effective solutions for construction projects and customers. The solutions include a performing of the graph represented generative floorplan, the scalable and user-participated framework. This research targets performing an end-to-end generative pipeline to conduct valid schematic floor plan designs for contextual adjustment representation of any functional buildings. The goals are also to develop a framework that accommodates the breadth of constraints necessary and makes the architectural design process efficient and expressive project. The methodology includes given the data collection in previous development, the graph learning and transformer approach will be utilized to represent the properties of rooms and their connectivity that extends the existing work into molecule generation; perform a scalable graph neural network and a transformer-based model to generative the floorplan; and developing an interactive, human-in-the-loop design tool framework for the architects.

Faculty Supervisor:

Yan Liu

Student:

Partner:

Maket technologies inc.

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

Concordia University

Program: