FloorPlan-DeepSeek (FPDS): A multimodal approach to floorplan generation using vector-based next room prediction

2025-06-11

arXiv.org

AI Summary

Existing floorplan generation models predominantly adopt end-to-end pixel-level synthesis, failing to align with architects’ incremental, iterative design workflows. Method: We propose a vectorized “next-room prediction” paradigm—the first to adapt autoregressive modeling from large language models to architectural layout generation—enabling structure-aware, incremental floorplan synthesis. Our approach employs a multimodal Transformer architecture that jointly encodes textual prompts and sequences of vectorized room representations, autoregressively generating parameterized rooms (type, dimensions, position, adjacency) while explicitly enforcing geometric and topological constraints. Results: On text-to-floorplan generation, our method matches or exceeds the performance of diffusion-based models and Tell2Design, achieving significant improvements in topological validity, functional coherence, and interactive controllability. This work establishes a new, practice-aligned paradigm for AI-assisted architectural design.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Natural Language Processing: Generation
• Humans and AI: Game Design — Procedural Content Generation & Storytelling

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

In the architectural design process, floor plan generation is inherently progressive and iterative. However, existing generative models for floor plans are predominantly end-to-end generation that produce an entire pixel-based layout in a single pass. This paradigm is often incompatible with the incremental workflows observed in real-world architectural practice. To address this issue, we draw inspiration from the autoregressive'next token prediction'mechanism commonly used in large language models, and propose a novel'next room prediction'paradigm tailored to architectural floor plan modeling. Experimental evaluation indicates that FPDS demonstrates competitive performance in comparison to diffusion models and Tell2Design in the text-to-floorplan task, indicating its potential applicability in supporting future intelligent architectural design.

Problem

• Addresses incompatibility of end-to-end floorplan models with incremental workflows
• Proposes next room prediction for progressive architectural design
• Enhances text-to-floorplan task performance compared to existing models

Innovation

• Vector-based next room prediction paradigm
• Multimodal approach for floorplan generation
• Autoregressive mechanism inspired by language models