DecoMind: A Generative AI System for Personalized Interior Design Layouts

Reema Alshehri, Rawan Alotaibi, Leen Almasri, Rawan Altaweel · 2025

Read DecoMind: A Generative AI System for Personalized Interior Design Layouts by Reema Alshehri, Rawan Alotaibi, Leen Almasri, Rawan Altaweel (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-16696.txt.

Introduction

Interior design layouts can be generated from simple user preferences, Reema Alshehri and co-authors present in a 2025 paper. Their system is called DecoMind. The user supplies inputs such as room type, style, and furniture preferences. From those plain inputs, the system builds a personalized layout and renders a realistic design image that a non-expert can judge at a glance.

A vision-language model extracts relevant furniture from a data set. The layout and a prompt feed a diffusion image model with structural control, which generates a design incorporating the selected furniture. Classifiers then evaluate the result to ensure alignment with the user's inputs. The pipeline offers an automated solution for realistic interior design, from preference list to picture.

Catalog record

CREATORReema Alshehri, Rawan Alotaibi, Leen Almasri, Rawan Altaweel
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE17 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-16696

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

DecoMind: A Generative AI System for Personalized Interior Design Layouts Reema Alshehri, Rawan Alotaibi, Leen Almasri, Rawan Altaweel Abstract This paper introduces a system for generating interior design layouts based on user inputs, such as room type, style, and furniture preferences. CLIP extracts relevant furniture from a dataset, and a layout that contains furniture and a prompt are fed to the Stable Diffusion with ControlNet to generate a design that incorporates the selected furniture. The design is then evaluated by classifiers to ensure alignment with the user’s inputs, offering an automated solution for realistic interior design. I. Introduction Interior design has become increasingly popular as people seek more comfort and personalization in their living spaces. While hiring professional designers is common for full-home projects, redesigning a single room—such as a bedroom—may not justify the cost or effort involved in hiring such services.Additionally, many individuals who prefer to furnish their rooms using items from specific stores like IKEA often feel uncertain about whether suggested furniture—based on their selected categories (e.g., sofa, table)—will suit the room’s size, layout, and style. This lack of visual clarity often leads to hesitation and indecision.Both of these challenges—cost and uncertainty—are addressed by our proposed system, DecoMind. The name combines “Deco” (short for decoration) and “Mind,” reflecting the system’s goal of enabling smart, AI-assisted design decisions. By entering basic information such as room dimensions, type, preferred style, and furniture categories, users can visualize how recommended furniture—currently supporting only IKEA items—might appear in their space. DecoMind is intended primarily to assist individuals redesigning a single room or validating design choices before purchase, complementing rather than replacing professional interior designers. II. Related Work Several previous research papers have explored AI-driven interior design generators. The CLIP-Layout model [ 1 ] introduced a method that learns visual correspondences between layout and text, enabling style-consistent and aesthetically pleasing scenes. However, this approach primarily focused on matching visual style and lacked structured control over furniture placement. Another approach is DeepFurniture [ 2 ] , which uses object detection and re-ranking modules to identify furniture in room images and recommend compatible sets from a fixed dataset. Although useful for retrieval, it does not support generating new layouts or adapting to customized user constraints such as room size or door/window placement. More recently, diffusion-based techniques have shown promise in generating realistic interior scenes. For example, CreativeDiffusion [ 3 ] integrates ControlNet with Stable Diffusion to guide image generation based on structural layouts. This model improves coherence between generated images and the room’s physical structure, yet its focus is limited to visual realism and does not include user-specific furniture or interactive customization. Our Contribution: Unlike previous works, our project DecoMind offers a fully integrated system that: • Combines CLIP and Stable Diffusion for furniture-aware generation based on both spatial structure and user preferences. • Allows users to define the room dimensions, door/window positions, furniture source (e.g., IKEA), and specific aesthetic styles. • Automatically extracts suitable furniture, filters complex scenes, removes backgrounds, and generates structured layout images. • Uses ControlNet-guided diffusion to synthesize realistic designs while preserving layout alignment. • Classifies the generated output and evaluates it against user intent, providing feedback for future tuning. This holistic approach bridges the gap between retrieval-based systems like DeepFurniture and style-guided models like CLIP-Layout, enabling a more personalized, interactive, and layout-consistent design experience. III. Data Description In developing the DecoMind system, we relied on three distinct datasets, given the lack of a unified dataset that encompasses room types, interior design styles, and the furniture associated with the store selected by the user for the generation process. As outlined in the introduction, the current version of our project is tailored exclusively to furniture available from the IKEA store. The datasets utilized are summarized as follows: • IKEA Furniture Dataset [ 4 ] : This dataset contains three folders: Train, Validation, and Test. Each folder includes images of IKEA furniture items (e.g., sofas, chairs, tables, etc.). It is primarily used to guide the model in generating images that feature IKEA furniture accurately. Additionally, we performed a custom filtering process to remove low-quality or irrelevant images, ensuring a cleaner and more accurate dataset for furniture representation. • House Rooms Image Dataset [ 5 ] : This dataset is organized into multiple folders, with each folder representing a specific room type (e.g., bedrooms, kitchens, living rooms, etc.). It enables the model to recognize and differentiate between various types of rooms. • Interior Design Styles Dataset [ 6 ] : This dataset consists of several folders, each corresponding to a particular interior style (e.g., modern, classic, minimalist, etc.). It helps the model identify and distinguish between various interior design styles. *Notes:The Interior Design Styles dataset was obtained from Kaggle, and it contains images originally sourced from Houzz.com. These images are subject to copyright restrictions. Their use is limited to academic and research purposes only, and redistribution is not allowed. IV. Methodology We have multiple stages in our study, each utilizing different methods for achieving the desired results. The following diagram provides an overview of the entire system architecture, illustrating the main components and the flow between them: Fig. 1 : Overview of the general workflow of the DecoMind system. As shown in the diagram, the system is composed of several interconnected modules, starting from data preparation to image generation and furniture selection. Each component plays a specific role in ensuring the final design aligns with the user’s input.A detailed explanation of each stage is provided below: A. Furniture Extraction: The process begins with extracting furniture from the selected store. The extraction process is driven by user choices, which include room type, style, and furniture types. This is achieved using a trained CLIP model, which uses a dual-encoder architecture—one for processing images and the other for processing text. CLIP matches the provided text (room type, style, furniture types) with relevant images to ensure the best outputs. Fig. 2 : Extracting furniture. B. Image Generation Using Stable Diffusion: The next stage involves the main component of the system: the Stable Diffusion model. This model generates images from text inputs using a latent diffusion process. It has achieved impressive results in recent years but has limitations, particularly in using reference images. In our case, we wanted the model to generate interior designs with the furniture selected by the user. To overcome this limitation, we integrated ControlNet, which guides Stable Diffusion using additional input such as the layout of the room (room dimensions, window and door positions) and the extracted furniture from the previous stage. -Stable Diffusion Architecture: Stable Diffusion uses a deep convolutional neural network with a transformer-based architecture for image generation. It operates by mapping the text prompt into a latent space, where it iteratively refines the image through a denoising process. -ControlNet Architecture: ControlNet acts as a conditioning network, guiding the generative process by adding constraints from the layout and furniture data, ensuring that the generated design adheres to the required dimensions and object placements. Fig. 3 : Generate the layout. After extracting the furniture and generating the room layout, the subsequent step involved constructing a descriptive text prompt for the image generation model.During this process, we faced limitations related to the IKEA dataset (store dataset),

Read the complete text