Abstract The growing need for rapid and responsive spatial planning solutions has created the demand for computational techniques that translate user intent into logical architectural configurations. This work presents a hybrid workflow integrating adjacency-based spatial reasoning with deep generative modelling for first-stage floor plan design. The framework consists of two interconnected modules: an interactive web platform that allows users to define room adjacencies, privacy zones, and sizes using an interactive matrix, and a Pix2Pix conditional GAN that translates such symbolic or sketch inputs into architectural floor plans. A scoring system evaluates adjacent correctness, room size deviations, privacy consistency, and circulation efficiency, providing automated feedback and suggestions for improvements. To validate the system, a case scenario was established using a controlled user study with three types of users-professional architects, students, and general users-each of which performed design tasks in two iterations, both unguided and guided by system recommendation. Findings presented quantifiable gains in adjacency fidelity, space efficiency, and final layout quality for all three groups, with the general users experiencing the biggest relative improvements. This project suggests how the combination of symbolic and generative reasoning can bridge the divide between abstract spatial comprehension and representation in architecture. The case study suggests the worth of the platform as a cooperative iteration tool for expert checking and novice acquisition in early-stage architectural design.
Rezve et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: