building floor plan graph representation

No abstract available

Nowadays, a lot of old floor plans exist in printed form or are stored as scanned raster images. Slight rotations or shifts may occur during scanning. Bringing floor plans of this form into a machine readable form to enable further use, still poses a problem. Therefore, we propose an end-to-end pipeline that pre-processes the image and leverages a novel approach to create a region adjacency graph (RAG) from the pre-processed image and predict its nodes. By incorporating normalization steps into the RAG feature extraction, we significantly improved the rotation invariance of the RAG feature calculation. Moreover, applying our method leads to an improved F1 score and IoU on rotated data. Furthermore, we proposed a wall splitting algorithm for partitioning walls into segments associated with the corresponding rooms.

We present our approach to improve room classification task on floor plan maps of buildings by representing floor plans as undirected graphs and leveraging graph neural networks to predict the room categories. Rooms in the floor plans are represented as nodes in the graph with edges representing their adjacency in the map. We experiment with House-GAN dataset that consists of floor plan maps in vector format and train multilayer perceptron and graph neural networks. Our results show that graph neural networks, specifically GraphSAGE and Topology Adaptive GCN were able to achieve accuracy of 80% and 81% respectively outperforming baseline multilayer perceptron by more than 15% margin.

Access graphs that indicate adjacency relationships from the perspective of flow lines of rooms are extracted automatically from a large number of floor plan images of a family-oriented rental apartment complex in Osaka Prefecture, Japan, based on a recently proposed access graph extraction method with slight modifications. We define and implement a graph convolutional network (GCN) for access graphs and propose a model to estimate the real estate value of access graphs as the floor plan value. The model, which includes the floor plan value and hedonic method using other general explanatory variables, is used to estimate rents, and their estimation accuracies are compared. In addition, the features of the floor plan that explain the rent are analyzed from the learned convolution network. The results show that the proposed method significantly improves the accuracy of rent estimation compared to that of conventional models, and it is possible to understand the specific spatial configuration rules that influence the value of a floor plan by analyzing the learned GCN.

This paper presents a new framework to classify floor plan elements and represent them in a vector format. Unlike existing approaches using image-based learning frameworks as the first step to segment the image pixels, we first convert the input floor plan image into vector data and utilize a graph neural network. Our framework consists of three steps. (1) image pre-processing and vectorization of the floor plan image; (2) region adjacency graph conversion; and (3) the graph neural network on converted floor plan graphs. Our approach is able to capture different types of indoor elements including basic elements, such as walls, doors, and symbols, as well as spatial elements, such as rooms and corridors. In addition, the proposed method can also detect element shapes. Experimental results show that our framework can classify indoor elements with an F1 score of 95%, with scale and rotation invariance. Furthermore, we propose a new graph neural network model that takes the distance between nodes into account, which is a valuable feature of spatial network data.

No abstract available

Abstract The generation of floor plan layouts has been extensively studied in recent years, driven by the need for efficient and functional architectural designs. Despite significant advancements, existing methods often face limitations when dealing with specific input adjacency graphs or room shapes and boundary layouts. When adjacency graphs contain separating triangles, the floor plan must include rectilinear rooms (non-rectangular rooms with concave corners). From a design perspective, minimizing corners or bends in rooms is crucial for functionality and aesthetics. In this article, we present a Python-based application called G-Drawer for automatically generating floor plans with a minimum number of bends. G-Drawer takes any plane triangulated graph as an input and outputs a floor plan layout with minimum bends. It prioritizes generating a rectangular floor plan (RFP); if an RFP is not feasible, it then generates an orthogonal floor plan or an irregular floor plan. G-Drawer modifies orthogonal drawing techniques based on flow networks and applies them on the dual graph of a given PTG to generate the required floor plans. The results of this article demonstrate the efficacy of G-Drawer in creating efficient floor plans. However, in future, we need to work on generating multiple dimensioned floor plans having non-rectangular rooms as well as non-rectangular boundary. These enhancements will address both mathematical and architectural challenges, advancing the automated generation of floor plans toward more practical and versatile applications.

This paper introduces a novel AI-based framework for automated 2D architectural floor plan generation from natural language descriptions, employing graph-conditioned denoising diffusion probabilistic models (DDPM) integrated with a U-Net architecture. The proposed system transforms user-provided textual specifications of room types, counts, and spatial relationships into structured room-adjacency graphs. These semantic graphs are encoded into compact embeddings and used to condition the diffusion model during the generative process. The framework produces segmented floor plans with clearly labelled architectural components, which are further refined through post-processing into editable 2D layouts. We evaluate our method on the RPLAN dataset and incorporate qualitative assessments from domain experts. Results demonstrate superior performance in generating diverse, semantically valid, and spatially coherent floor plans compared to existing GAN- and transformer-based approaches. This work highlights the effectiveness of combining semantic graph representations with diffusion-based generative models in advancing AI-driven architectural design.

Let G = (V, E) be a maximal planar graph (MPG), where every face is triangular. A floor plan (FP) of an n-vertex MPG G is a partition of a rectangle into n rectilinear polygons called modules where two modules are adjacent if and only if there is an edge between the corresponding vertices in G. It can be easily found that it is not possible to construct a FP for a given MPG while maintaining the rectangularity of the modules of a FP (for an example, consider the complete graph K4). Hence, to satisfy adjacency requirements of a MPG, bends need to be introduced within the FPs, where a bend is a concave corner of a module in a FP. A FP with rectilinear modules or with at least one bend is called orthogonal floor plan (OFP). There exist algorithms for the construction of an OFP for a given MPG but the notion of minimum bends within an OFP is not yet discussed in the literature. In this paper, a mathematical procedure for computing the minimum number of bends required in an OFP for a MPG G has been presented. Further, it has been shown that the number of bends in an OFP depends only on critical separating triangles and K4’s.

No abstract available

No abstract available

No abstract available

This study explores enhancing spatial analysis by integrating Space Syntax with deep learning methods for images and graphs. While Space Syntax traditionally focuses on geometry and topology, it overlooks visual data like texture and color. As image content varies by location, capturing spatial context is essential, especially in irregular or open spaces where identifying central points is challenging. To address this, the study introduces the concept of an “isovist graph” - a spatial model that minimally covers a closed plane with isovists while preserving centrality and connectivity. A method is proposed to compute this model rigorously in discrete settings. Experiments show that exact solutions can be efficiently derived for many spatial cases, aligning with the intended objectives.

No abstract available

We address the challenging problem of floor plan reconstruction from sparse views and a room-connectivity graph. As a first stage, we construct a flexible graph-structure unifying the connectivity graph and the sparse observed data. Using our Graph Neural Network architecture, we can then refine the available information and predict unobserved room properties. In a second step, we introduce a Constrained Diffusion Model to reconstruct consistent floor plan matching the available information, despite of its sparsity. More precisely, we use a Cross-Attention mechanism armed with shape descriptors to guarantee that the generated floor plan reflects both the input room connectivity and the geometry observed in the sparse views.

ABSTRACT Floor planning, as one of the most important considerations in building design, often involves intensive trial-and-error processes with many constraints considered simultaneously. Artificial intelligence (AI) generative design solutions being developed are hampered by two shortcomings. Firstly, the vast topological knowledge embedded in existing floor plans has been largely unexplored and is thus wasted. Secondly, an efficient methodological instrument to learn the topological knowledge for generative design is lacking. This paper aims to develop a graph learning methodology to learn graph knowledge from floor plans and generate knowledge ready for building generative design. A discrete denoising diffusion model (D3M) that can learn topology graphs via its bi-directional structure of ‘corruption and denoise’ is developed and trained using more than 80,000 floor plans from a large-scale dataset. It is found that the D3M can learn the knowledge from floor plans and present it as various building floor topologies, which are evaluated in a preliminary case study as reliable and useful for generating real-life building floor plans. The research provides a design knowledge management framework that can be further implemented in academic works and design practices through some mainstreaming or commercializing efforts.

We propose a simple yet effective metric that measures structural similarity between visual instances of architectural floor plans, without the need for learning. Qualitatively, our experiments show that the retrieval results are similar to deeply learned methods. Effectively comparing instances of floor plan data is paramount to the success of machine understanding of floor plan data, including the assessment of floor plan generative models and floor plan recommendation systems. Comparing visual floor plan images goes beyond a sole pixel-wise visual examination and is crucially about similarities and differences in the shapes and relations between subdivisions that compose the layout. Currently, deep metric learning approaches are used to learn a pair-wise vector representation space that closely mimics the structural similarity, in which the models are trained on similarity labels that are obtained by Intersection-over-Union (IoU). To compensate for the lack of structural awareness in IoU, graph-based approaches such as Graph Matching Networks (GMNs) are used, which require pairwise inference for comparing data instances, making GMNs less practical for retrieval applications. In this paper, an effective evaluation metric for judging the structural similarity of floor plans, coined SSIG (Structural Similarity by IoU and GED), is proposed based on both image and graph distances. In addition, an efficient algorithm is developed that uses SSIG to rank a large-scale floor plan database. Code will be openly available.

In this paper, we present a GNN-based Line Segment Parser (GLSP), which uses a junction heatmap to predict line segments' endpoints, and graph neural networks to extract line segments and their categories. Different from previous floor plan recognition methods, which rely on semantic segmentation, our proposed method is able to output vectorized line segment and requires less post-processing steps to be put into practical use. Our experiments show that the methods outperform state-of-the-art line segment detection models on multi-class line segment detection tasks with floor plan images. In the paper, we use our floor plan dataset named Large-scale Residential Floor Plan data (LRFP). The dataset contains a total of 271,035 floor plan images. The label corresponding to each picture contains the scale information, the categories and outlines of rooms, and the endpoint positions of line segments such as doors, windows, and walls. Our augmentation method makes the dataset adaptable to the drawing styles of as many countries and regions as possible.

No abstract available

No abstract available

In recent years, machine learning has been widely used in the real estate field. However, most of these previous studies have been limited to analysis based on objective perspectives, such as analysis of the structure of the floor plan and rent estimation. On the other hand, we focus on the subjective "living comfort" of real estate properties and aim to predict people's impressions of properties based on information obtained from floor plan images. Specifically, by using deep learning to analyze floor plan images and graph structures reflecting the floor plans, it becomes possible to predict the attractiveness of each property in terms of spaciousness, modernity, privacy, and so on. As a result of the experiments, the effectiveness of using both the floor plan image and the corresponding graph structure for prediction was confirmed.

No abstract available

Floor plan images describe the layout of real estates. Although they provide essential information on properties for evaluating their value, they are intractable for computers due to various drawing styles owing to lack of standard formats. The purpose of this paper is to automatically transform a floor plan image into a graph that is well-structured and mathematically well-defined representation. This approach facilitates comparing and evaluating floor plans and can be applied to many systems handling floor plans. Our method achieves it with 92% accuracy. Furthermore, we show similar property retrieval using the extracted graphs as an example of possible applications.

To overcome this problem of floor plan recognition, a new hybrid architecture Centrality-Based GraphSAGE (CB-SAGE) is presented which models the problem as a graph classification problem. The actual building architectures are translated into a graph structure with each room/compartment becoming a node and the interconnectivity among rooms becoming the graph edges. However, the regular GraphSAGE methods mostly depend on the local neighbourhood sampling and aggregation, risking to neglect more structural properties. CB-SAGE improve upon this by embedding centrality-based measures like clustering coefficient or betweenness centrality directly into the feature matrix of nodes during training. Such metrics provide the graph embeddings with topological context that allows the model to learn more discriminative and informative pattern. Due to that, the model can capture more unique spatial plans and categorizing sophisticated types of floor plans. Evaluation on real-world datasets indicate that CB-SAGE considerably gains the standard models in regard to accuracy and generalization capacity. The approach is of practical value to the fields of architectural analysis, robotics navigation, and the design of smart buildings, where the effective and meaningful graph-based spatial environment representation is important. CB-SAGE unites the benefits of local aggregation with the global structural information, thus providing a potent new paradigm of graph based spatial learning. The proposed system had improved results over alternative approaches and recorded high Macro-F1 score of 77.5 % and purity of 75.4 % on the House-GAN floor plan dataset. It was always outperforming GCN, GAT, and GraphSAGE.

In this paper, we propose an end-end model for producing furniture layout for interior scene synthesis from the random vector. This proposed model is aimed to support professional interior designers to produce the interior decoration solutions more quickly. The proposed model combines a conditional floor-plan module of the room, a conditional graphical floor-plan module of the room and a conditional layout module. As compared with the prior work on scene synthesis, our proposed three modules enhance the ability of auto-layout generation given the dimensional category of the room. We conduct our experiments on the proposed real-world interior layout dataset that contains $191208$ designs from the professional designers. Our numerical results demonstrate that the proposed model yields higher-quality layouts in comparison with the state-of-the-art model. The dataset and code are released \href{this https URL}{Dataset,Code}

We introduce ResPlan, a large-scale dataset of 17,000 detailed, structurally rich, and realistic residential floor plans, created to advance spatial AI research. Each plan includes precise annotations of architectural elements (walls, doors, windows, balconies) and functional spaces (such as kitchens, bedrooms, and bathrooms). ResPlan addresses key limitations of existing datasets such as RPLAN (Wu et al., 2019) and MSD (van Engelenburg et al., 2024) by offering enhanced visual fidelity and greater structural diversity, reflecting realistic and non-idealized residential layouts. Designed as a versatile, general-purpose resource, ResPlan supports a wide range of applications including robotics, reinforcement learning, generative AI, virtual and augmented reality, simulations, and game development. Plans are provided in both geometric and graph-based formats, enabling direct integration into simulation engines and fast 3D conversion. A key contribution is an open-source pipeline for geometry cleaning, alignment, and annotation refinement. Additionally, ResPlan includes structured representations of room connectivity, supporting graph-based spatial reasoning tasks. Finally, we present comparative analyses with existing benchmarks and outline several open benchmark tasks enabled by ResPlan. Ultimately, ResPlan offers a significant advance in scale, realism, and usability, providing a robust foundation for developing and benchmarking next-generation spatial intelligence systems.

Indoor maps are essential for navigation, resource allocation, and autonomous operation in complex environments, yet creating them at scale has long been impeded by high costs and specialized hardware requirements. We present Tesseract, a modular system that transforms ordinary low-semantic floor plan images into navigable graph structures, without requiring specialized sensors or 3D modeling tools. Through Tesseract, we integrate deep learning modules for text detection and door classification. We then implement a novel floodfill-based segmentation and graph optimization solution. Tesseract ultimately generates semantically rich, compact graph representations of the original floor plans that are computationally parsable for indoor navigation applications. We evaluate Tesseract across two large-scale university buildings as well as a benchmark dataset, demonstrating high navigational completeness despite variations in layout complexity. The system processes floor plans efficiently, with runtime scaling linearly to the number of detected regions, thus remaining practical for large-scale deployments. Graph pruning reduces the initially dense connectivity—typically quadratic in the number of regions—to a sparse structure, yielding up to 78% fewer nodes and 70% fewer edges, all without compromising connectivity. Moreover, geometric fidelity is preserved within 80–86% of true real-world distances. These findings establish Tesseract as a robust and scalable solution, broadening access to automated indoor navigation and spatial analytics.

: We present a practical and efficient approach to residential architectural space planning. Our method allows users to input the house boundaries and their desired room constraints, and the application then assists in generating a well-designed floor plan. By leveraging machine learning-based algorithms, our application effectively partitions the house boundaries into individual rooms, resulting in an optimized layout that adheres to the user's specified constraints and preferences. In this approach, we integrate the PLANFORGE module, which is inspired by the Graph2Plan model, to generate the floor plan. This module transforms the user-provided boundary data into a structured layout graph. This enables us to provide personalized solutions that align with the user's desired boundaries. So, our system automatically allocates and refines the room divisions, ensuring a harmonious arrangement of the available space.

No abstract available

No abstract available

Abstract: This paper presents an optimized method for reconstructing 3D floor plans using userdefined boundaries and constraints for residential structures. This approach allows users to providearchitectural constraints such as room types and quantities, as well as manual sketches or existingimages of the house boundaries. Advanced deep learning algorithms are used to automaticallypartition the house boundaries and create a customized, optimized interior layout based on the users’architectural constraints. In the experiment phase, we integrated the Graph2Plan-based deep learningmodule, which converts the user-provided boundary data and architectural constraints into astructured 2D floor plan, automatically allocating and refining the rooms to ensure a harmoniousspatial arrangement. The evaluation of the deep learning model’s performance shows that this is auseful and time-saving solution for designers. Then, we utilized graphics and image processingtechniques to generate the 3D floor plans. Based on this solution, we have developed a 3D floor plangeneration application that provides a flexible and adaptive solution for individual home planningwithin defined boundaries. The application has been thoroughly tested to demonstrate its features,including the ability to meet users’ architectural constraints, provide rapid response times, and offera convenient user interaction experience.Keywords: 3DFloorplan, Floorplan generation, layout graph, RPLAN dataset, house plan.

Floor plans depict building layouts and are often represented as graphs to capture the underlying spatial relationships. Comparison of these graphs is critical for applications like search, clustering, and data visualization. The most successful methods to compare graphs \ie, graph matching networks, rely on costly intermediate cross-graph node-level interactions, therefore being slow in inference time. We introduce \textbf{LayoutGKN}, a more efficient approach that postpones the cross-graph node-level interactions to the end of the joint embedding architecture. We do so by using a differentiable graph kernel as a distance function on the final learned node-level embeddings. We show that LayoutGKN computes similarity comparably or better than graph matching networks while significantly increasing the speed. \href{https://github.com/caspervanengelenburg/LayoutGKN}{Code and data} are open.

No abstract available

Existing floor plan retrieval methods match residential floor plans based on room function and adjacency, ignoring the shape of the interior rooms. Inspired by shape grammars, we incorporate interior layout into the similarity metric using a tree structure that represents both the layout hierarchy and room shapes of the floor plan. We create parse trees from floor plans and evaluate their similarity using an appropriately defined tree edit distance. We evaluate the method on a public dataset of 11,250 vector graphic representations of Japanese homes. A user study shows that our method retrieves layouts preferential to those obtained using the exterior outline or room adjacency matrix.

3D residential reconstruction plays a crucial role in various fields, such as architecture, interior design, virtual reality, and robot navigation. While existing methods typically rely on complex data acquisition setups or multiple images, we proposed a novel method for 3D residential reconstruction from a single indoor floor plan. By leveraging the information encoded in an indoor floor plan, our proposed method utilizes traditional and deep-learning-based computer vision techniques to reconstruct the corresponding 3D residential model accurately. Image classification based on deep learning is applied for calculating the scale, and semantic segmentation based on deep learning is employed for extracting walls, doors and windows. By analyzing the complex geometric relationship of skeleton curves, the vectorized representation of the building structure is obtained. An accurate 3D residential model is eventually reconstructed by synthesizing all the obtained information.

Following the previous paper where an evacuation time calculation algorithm was implemented based on manually prepared data, this one formalizes a final end-to-end module for importing a PDF/image evacuation plan and automatically converting it into an object and graph model for further calculations according to DSTU standards. A hybrid Computer Vision pipeline is pro-posed: vector analysis of PDF objects (lines/polylines/texts/layers); raster branch using OpenCV (HSV segmentation of green «EXIT» signs and arrows, wall detection using LSD/Hough, door detection as local «gaps», morphological skeletonization, and distance trans-form for estimating the local passage width). The module output consists of a set of objects (ex-its, doors, stairs, sections, passages) and a directed graph (V, E) with attributes of length and width, directly consumed by the software from the first paper [1]. To evaluate the performance of automatic evacuation plan import, several key factors are considered. First, we check wheth-er the system correctly finds the exits. Next, the accuracy of recognized directional arrows, which indicate movement orientation, is assessed. For corridors, it is important that the gener-ated «passage mask» matches the actual layout. Additionally, the connectivity of the graph is analyzed — whether it is possible to reach an exit from any point — and whether the routes can be fully reconstructed. Moreover, the final outcome, the calculated evacuation time, is evaluat-ed and compared with manual calculations according to the DSTU methodology. Thus, both the correctness of object recognition and the accuracy of the final calculations are verified. The re-sult is the elimination of manual input to the program, reproducibility, and suitability for re-porting. The module is integrated with the previously developed object-oriented software (clas-ses Area, EvacuationRoute, Metrics; BFS/DFS), ensuring a full cycle.

With the advent of AI and computer vision techniques, the quest for automated and efficient floor plan designs has gained momentum. This paper presents a novel approach using skip-connected neural networks integrated with layout graphs. The skip-connected layers capture multi-scale floor plan information, and the encoder-decoder networks with GNN facilitate pixel-level probability-based generation. Validated on the MSD dataset, our approach achieved a 93.9 mIoU score in the 1st CVAAD workshop challenge. Code and pre-trained models are publicly available at https://github.com/yuntaeJ/SkipNet-FloorPlanGe.

No abstract available

In the context of floorplan generation, a recent trend has emerged towards incorporating user-provided descriptive texts to deliver a more flexible and user-centric interface, en-abling direct user engagement in the design and configuration processes. However, extracting accurate information from text input to characterize corresponding floorplans is challenging due to the latent representation of semantic, geometric, and topological design details. Conversely, graph-based generation methods, while lacking geometric information, exhibit superior performance owing to the intuitive representation of spatial data in graph structures. Building upon this understanding, this paper proposes a dual-stage framework that augments text input with serialized graphs to leverage the complementary strengths of both modalities. The pre-training stage uses script-generated texts to empower a graph generator for producing corresponding graphs from descriptive texts and familiarize a floorplan generator with concatenated text and graph input. During the fine-tuning stage, the pre-trained graph generator, alongside an alternative graph generator based on prompt-driven large language models, is employed to provide augmented graphical knowledge for the floorplan generator based on manually annotated descriptive texts. Extensive quantitative and qualitative experiments demon-strate the effectiveness of the proposed approach over the vanilla text- based floorplan generation method.

Effective navigation in unfamiliar environments remains a critical challenge for successful deployment. Current navigation methods, which rely on autonomous or teleoperated exploration and map building, pose technical difficulties for inexperienced end-users. In contrast, humans can effectively navigate using abstract floorplans, suggesting the potential for service robots to leverage similar techniques. The practical application of floorplan-based navigation; however, is currently limited by methods that require pre-exploration or floorplan with accurate measurements or scale. This letter aims to address the aforementioned challenges and investigate the feasibility of floorplan-based navigation in unfamiliar environments. Specifically, we propose a novel scale-invariant floorplan localization method, enabling navigation without relying on precise scale information. Furthermore, we introduce an incremental graph augmentation approach that enriches the floorplan representation with traversability information derived from robot observations. Finally, we develop an efficient navigation framework capable of utilizing both the inherent structure of the floorplan and real-time observations. Experimental results demonstrate that our scale-invariant floorplan localization method outperforms baseline methods in most cases when floorplan scale information is unavailable, and our graph-based navigation system exhibits superior success and efficiency as compared to grid-based counterparts. The outcomes of this research contribute to the advancement of service robot deployment in unfamiliar environments, particularly in scenarios where extensive exploration and map building may be impractical or technically challenging for end-users.

ABSTRACT Building patterns are important components of urban structures and functions, and their accurate recognition is the foundation of urban spatial analysis, cartographic generalization, and other tasks. Current building pattern recognition methods are often based on a shape index that can only characterize shape features from one aspect, resulting in significant errors. In this study, a building pattern recognition method based on a graph neural network is proposed to enhance shape cognition and focus on recognizing collinear patterns. First, a building shape classification model that integrates global shape and graph node structure features was constructed to quantitatively study shape cognition. Subsequently, a collinear pattern recognition (CPR) model was established based on a dual building graph. The shape cognition results were integrated into the model to enhance its recognition ability. The results show that the shape classification model can be used to effectively distinguish different shape categories and support building pattern recognition tasks. Based on the CPR model, false recognitions can be avoided, and recognition results similar to those of visual cognition can be obtained. Compared with the comparative methods, both models have significant advantages in terms of statistical results and implementation.

In recent times, researchers have proposed several approaches for building floorplans using parametric/generative design, shape grammars, machine learning, AI, etc. This paper aims to demonstrate a mathematical approach for the automated generation of floorplan layouts. Mathematical formulations warrant the fulfilment of all input user constraints, unlike the learning‐based methods present in the literature. Moreover, the algorithms illustrated in this paper are robust, scalable and highly efficient, generating thousands of floorplans in a few milliseconds.

Regularized vector representation of artificial buildings serves as foundation of remote sensing and GIS analysis tasks. To generate exact and regular building polygons, a graph convolutional network integrating spatial morphology cognition features is proposed. Abstract image semantic features from CNN and topological spatial-cognition features from graph reasoning module are united for node features in graph construction and dynamic convolutions. The proposed model is validated on WHU building dataset and compared with Mask R-CNN and Curve-GCN. The superior performances demonstrate that the spatial cognitive information perception strategy enhances the reasonable geometric prediction on coordinates of boundary vertices.

Vector graphics (VG) are ubiquitous in industrial designs. In this paper, we address semantic segmentation of a typical VG, i.e., roughcast floorplans with bare wall structures, whose output can be directly used for further applications like interior furnishing and room space modeling. Previous semantic segmentation works mostly process well-decorated floorplans in raster images and usually yield aliased boundaries and outlier fragments in segmented rooms, due to pixel-level segmentation that ignores the regular elements (e.g. line segments) in vector floor-plans. To overcome these issues, we propose to fully utilize the regular elements in vector floorplans for more integral segmentation. Our pipeline predicts room segmentation from vector floorplans by dually classifying line segments as room boundaries, and regions partitioned by line segments as room segments. To fully exploit the structural relationships between lines and regions, we use two-stream graph neural networks to process the line segments and partitioned regions respectively, and devise a novel modulated graph attention layer to fuse the heterogeneous information from one stream to the other. Extensive experiments show that by directly operating on vector floorplans, we outper-form image-based methods in both mIoU and mAcc. In addition, we propose a new metric that captures room integrity and boundary regularity, which confirms that our method produces much more regular segmentations. Source code is available at https://github.com/DrZiji/VecFloorSeg.

Various factors are considered when designing a floorplan layout, including the plan’s outer boundary, room shape and size, adjacency, privacy, and circulation space, among others. While graph-theoretic approaches have proven effective for floorplan generation, existing algorithms generally focus on defining the boundary of the plan or different room shapes, lacking the investigation of designing circulation space within a floorplan. However, the circulation design in architectural planning is a crucial factor that affects the functionality and efficiency of areas within a building. This paper presents a graph-theoretic approach for integrating circulation within a floorplan. In this study, we use plane graphs to represent floorplans and develop graph algorithms to incorporate various types of circulation within a floorplan as follows: i. The first phase generates a spanning circulation, that is, a corridor leading to each room using a circulation graph. ii. Subsequently, using an approximation algorithm, the circulation space is minimized, that is, generation of minimum circulation space covering all the rooms, thereby enhancing space utilization in the floorplan. iii. Furthermore, customized circulations are generated to cater to user preferences, distinguishing between public and private spaces within the floorplan. In addition to the theoretical framework, we have implemented our algorithms in Python and developed a user-friendly graphical interface (GUI), enabling seamless integration of our algorithms into architectural design processes.

No abstract available

Indoor floorplans are widely used in fields like building information modeling, indoor navigation, emergency response, smart buildings, and architectural design simulation. However, reconstructing accurate floorplans from indoor laser point clouds is challenging due to the complexity, clutter, and occlusions of indoor structures. We propose kinetic expansion of linear structural elements (KELSE), an indoor scene floorplan reconstruction method to address these challenges. We design a structural element extraction method that integrates geometric feature constraints with semantic information to identify structural elements such as walls, doors, windows, ceilings, and floors in complex indoor scenes. A kinetic data structure expansion and undirected graph optimization are then used to reconstruct the complete floorplan. Experimental results show that KELSE achieves high accuracy and completeness, with room reconstruction reaching 0.98 and 0.95, respectively. KELSE provides an efficient and precise solution for floorplan reconstruction from indoor LiDAR point cloud data.

The use of neural networks to retrieve relevant images has become mainstream. However, retrieving images that contain specific spatial relationships remains a challenging task. Images alone are not sufficient to fully describe spatial and topological relationships, which are usually better represented as a graph made up of nodes and edges. This paper describes the development of a graph-based computational tool for retrieving architectural precedents that closely match the relationship between a building and its surrounding ground as detected in a designer’s project. The tool, titled Building Ground Relationship (BGR), stems from a research project into Graph Machine Learning (GML) that used Deep Graph Convolutional Neural Networks (DGCNNs) to classify building and ground relationships. The neural network was trained using a large synthetic dataset of graphs and optimized through the fine-tuning of its hyperparameters. To verify its performance, a second surrogate model was built using the Deep Graph Library (DGL). The results were nearly identical, thus giving confidence that the model is highly optimized. In the development of the BGR tool, two primary technologies were utilized. In the first instance, the synthetic database was built in Rhino Grasshopper by generating variations of a parametric model. The dual graphs of these models were then automatically generated and exported using the Toplogic software library. The second phase involved developing GML models used for predicting the class of the conceptual design, enabling the retrieval of the smaller case study. The results of this research point to the importance of topological representation and machine learning approaches in retrieving and classifying architectural precedents.

No abstract available

Deep learning (DL) models are now a reality towards the automation of the placement task of analog integrated circuit (IC) layout design, promising to bypass the limitations of existing approaches. However, as the complexity of analog design cases tackled by these methodologies increases, a broader set of topological constraints must be supported to cover different layout styles and circuit classes. Here, model-independent differentiable encodings for regularity, boundary, and symmetry island (SI) constraints are described, and an unsupervised loss function is used for the artificial neural network (ANN) model to learn how to generate placements that follow them. As only sizing data is required for its training, it discards the need to acquire legacy layouts containing insights of these types of constraints. The model is ultimately used to produce floorplans from scratch, at push-button speed, for state-of-the-art analog structures, including technology nodes not used for its training.

Given a building floorplan, humans can localize themselves by matching the observation of the environment with the floorplan using geometric, semantic, and topological clues. Inspired by this insight, this paper proposes a learning- based topometric robot localization method FloorplanNet, which implements a match between a metric robot map and the potentially inaccurate building floorplan in nonuniform scales and different shapes by semantic information. The method uses a novel Graph Neural Network to learn descriptors of nodes from topometric graphs generated from the input maps. We demonstrate that our method can match the 3D point cloud sub-map generated by the robot during the SLAM process with the 2D map. Furthermore, we apply our map-matching algorithm for real-world robot localization. We evaluate our method on several publicly available real-world datasets. Even though our network is solely trained using simulation data, our method demonstrates high robustness and effectiveness in real- world indoor environments and outperforms the existing SOTA map-matching algorithms. We further develop a simulator that automatically creates and annotates the required training data to train our neural networks. The method and simulator are released at: https://github.com/fengdelin/FloorplanNet.git

We present a novel approach to localize an unknown planar sensor network based on sparse sampling of partially observable paths traversed by moving agents. The problem is inspired by mapping the geometry of a building floorplan via "uncooperative sensing", using data from camera feeds and other tracking-capable sensors. Unique challenges include having no knowledge of sensor placement, coverage or their extrinsic parameters nor the knowledge of the motion of the people within a floorplan. The methods used are, at first, topological, to build a combinatorial model with the appropriate topology. This model is then augmented to include weak geometric information, and optimization techniques are used to approximate the domain. Topological information is captured within the optimization problem to constrain the solution.

This paper presents a new method for chip floorplanning optimization using deep learning (DRL) combined with graph neural networks (GNNs). The plan addresses the challenges of traditional floor plans by applying AI to space design and intelligent space decisions. Three-head network architecture, including a policy network, cost network, and reconstruction head, is introduced to improve feature extraction and overall performance. GNNs are employed for state representation and feature extraction, enabling the capture of intricate topological information from chip netlists. A carefully designed reward function incorporating wire length minimization, area utilization, and timing constraint satisfaction guides the DRL agent toward high-quality floorplan solutions. An exploration bonus based on reconstruction error addresses the sparse reward problem. Extensive testing of the ISPD 2005 benchmarks demonstrated the effectiveness of the proposed approach, consistently operating on a state-of-the-art basis. Significant improvements include an average 31.4% reduction in half-perimeter wire length (HPWL) and a 34.2% reduction in breach time compared to the best baseline performance. The process scalability and robustness are evaluated, showing performance in various circuits and different perturbations. This research advances AI-driven electronic device design and paves the way for better chip design processes.

This paper presents an innovative approach toward the automation of the placement task of analog integrated circuit layout design by using an artificial neural network that generates multiple valid floorplan solutions at push-button speed. The proposed model extends the knowledge mining of the most recent layout generation techniques as an end-to-end approach. A novel loss function is used in a semi-supervised fashion for the model to learn how to generate effective placements that follow topological constraints instead of simply trying to copy devices' locations from some pre-existing labeled dataset of placement solutions. Thus, in addition to eliminating the need for a large dataset of placed sizing solutions for training, the model generalizes better and outputs better layouts for solutions outside the training set. Moreover, one step further is taken towards a model that can predict the placement of different circuit topologies by supporting different encodings (with different number of devices) on the input layer of the artificial neural networks, ultimately fostering an opportunity to reuse incomplete legacy layout information. As proof of concept, the one ANN trained for 2 amplifier topologies is used to produce multiple placement recommendations in less than 40 ms for new designs.

As the technology node shrinks to the nanometer regime, the demand for new lithography methods with high resolution and low cost is increasing. Electron beam lithography (EBL) is one of the promising next-generation lithography (NGL) technologies that can tackle both challenges compared to the traditional lithography methods. Fogging effect, a culprit of pattern distortion in layout, is one of the major challenges that prevent industry from adopting EBL in technologies below 22nm. This paper proposes a reinforcement-learning (RL) placement method that trains a neural network as an agent to effectively control fogging effect. To speed up our method, we benefit from the following innovations: using topological floorplan representation for the layouts during placement, and deploying a RL trained agent that can intelligently take actions. The experimental results show that our proposed placer is able to more effectively and efficiently reduce the fogging effect variation in the analog circuits in comparison with the conventional simulated-annealing-based placement method.

We present CAGE (Continuity-Aware edGE) network, a robust framework for reconstructing vector floorplans directly from point-cloud density maps. Traditional corner-based polygon representations are highly sensitive to noise and incomplete observations, often resulting in fragmented or implausible layouts.Recent line grouping methods leverage structural cues to improve robustness but still struggle to recover fine geometric details. To address these limitations,we propose a native edge-centric formulation, modeling each wall segment as a directed, geometrically continuous edge. This representation enables inference of coherent floorplan structures, ensuring watertight, topologically valid room boundaries while improving robustness and reducing artifacts. Towards this design, we develop a dual-query transformer decoder that integrates perturbed and latent queries within a denoising framework, which not only stabilizes optimization but also accelerates convergence. Extensive experiments on Structured3D and SceneCAD show that CAGE achieves state-of-the-art performance, with F1 scores of 99.1% (rooms), 91.7% (corners), and 89.3% (angles). The method also demonstrates strong cross-dataset generalization, underscoring the efficacy of our architectural innovations. Code and pretrained models are available on our project page: https://github.com/ee-Liu/CAGE.git.

The task of reconstructing indoor floorplans has become an increasingly popular subject, offering substantial benefits across various applications such as interior design, virtual reality, and robotics. Despite the growing interest, existing approaches frequently encounter challenges due to high computational costs and sensitivity to errors in primitive detection. In this article, we introduce PolyGraph, a new computational framework that combines a deep-learning based primitive detection network with an optimization-based reconstruction algorithm to facilitate high-quality reconstruction results. Specifically, we develop a novel guided wall point primitive estimation network capable of generating dense samples along wall boundaries. This network not only retains structural detail but also shows improved robustness in the detection phase. Then, PolyGraph utilizes wall points to establish a graph-based representation, formulating indoor floorplan reconstruction as a subgraph optimization problem. This approach significantly reduces the search space comparing to existing pixel-level optimization approaches. By utilizing “structural weight”, we seamlessly integrate the structural information of walls and rooms into graph representations, ensuring high-quality reconstruction results. Experimental results demonstrate PolyGraph's effectiveness and its advantages compared to other optimization-based approaches, showcasing its computational efficiency, and its ability to preserve structural integrity and capture fine details, as quantified by the structure metrics.

Recognizing the detailed information embedded in rasterized floorplans is at the research forefront in the community of computer graphics and vision. With the advent of deep neural networks, automatic floorplan recognition has made tremendous breakthroughs. However, co‐recognizing both the structures and semantics of floorplans through one neural network remains a significant challenge. In this paper, we introduce a novel framework Raster‐to‐Graph, which automatically achieves structural and semantic recognition of floorplans. We represent vectorized floorplans as structural graphs embedded with floorplan semantics, thus transforming the floorplan recognition task into a structural graph prediction problem. We design an autoregressive prediction framework using the neural network architecture of the visual attention Transformer, iteratively predicting the wall junctions and wall segments of floorplans in the order of graph traversal. Additionally, we propose a large‐scale floorplan dataset containing over 10,000 real‐world residential floorplans. Our autoregressive framework can automatically recognize the structures and semantics of floorplans. Extensive experiments demonstrate the effectiveness of our framework, showing significant improvements on all metrics. Qualitative and quantitative evaluations indicate that our framework outperforms existing state‐of‐the‐art methods. Code and dataset for this paper are available at: https://github.com/HSZVIS/Raster-to-Graph.

The field of floorplan generation has attracted significant interest from the community. Remarkably, recent advances in generative models have markedly enhanced the development of this field. However, generating floorplans that satisfy various conditions remains a challenging task. This paper proposes a learning framework, named Cons2Plan, for automatically and high-quality generating vector floorplans from various conditions. The input conditions can be graphs, boundaries, or a combination of both. The conditional diffusion model is the core component of our Cons2Plan. The denoising network uses a conditional embedding module to incorporate the conditions during the reverse process. Additionally, Cons2Plan incorporates a two-stage approach that generates graph conditions based on boundaries. It uses three networks for node prediction and a novel conditional edge generation diffusion model, named CEDM, for edge generation. We conduct qualitative evaluations, quantitative comparisons, and ablation studies to show that our method produces better floorplans than state-of-the-art methods.

Validating house construction plans for compliance with building regulations and guidelines is usually a manual and slow process. This may lead to long delays and inconsistent evaluations. This paper proposes ARCH (Automated Regulation Checker for Housing), an AI-powered platform to automate the validation of $2 D$ construction plans. The platform aims to reduce the review time and enhance stakeholder satisfaction. ARCH uses AI and rule-based validation to evaluate floor plans. It combines YOLOv8 for symbol detection, Raster-to-Graph model for floor plan vectorization, and EasyOCR for text extraction. ARCH was evaluated through a user study involving 31 participants. Over $90 \%$ of users agreed that the platform is easy to use through its integrated submission, review, and tracking features.

No abstract available

No abstract available

No abstract available

Floor plan restoration aims to recover vector and semantic information from raster floor plan images, which is significant for advanced applications including interior design, interative walkthroughs, and layout planning. Existing methods generally adopt a two-stage paradigm: a parsing stage to extract semantic elements such as rooms, walls, doors, and windows from raster images; and then a vectorization stage to convert them into vector graphics. However, these methods are deficient in both accuracy and efficiency due to the neglect of the unique cues of floor plans compared to natural images. To address the above issues, we propose MMParseNet that yields accurate parsing results by incorporating multimodal cues unique to floor plans, such as room names, furniture icons, and room boundaries. Moreover, we implement an efficiency-optimized vectorization method based on PCA that avoids unnecessary iterative solutions. We conduct both quantitative and qualitative experiments on three public and one self-built dataset. The results exhibit consistent improvements in accuracy and sub-second overall restoration time across various datasets.

The floor plan recognition and vectorization problem from the image has a high market response due to the ability to be applied in such domains as design, automatic furniture fitting, property cost estimation, etc. Several approaches already exist on the market. Many of them are using just statistical or deep machine learning methods capable of recognizing a limited set of floor plan types or providing a semi-automatic tool for recognition. This paper introduces the approach based on the combination of statistical image processing methods in a row of machine learning techniques that allow training robust model for the different floor plan topologies. Faster R-CNN for the floor object detection with a mean average precision of 86% and UNet for the wall segmentation has shown the IoU metric results of about 99%. Both methods, combined with functional and component filtration, made it possible to implement the new approach for vectoring the floor plans.

The boundary-constrained floor plan generation problem aims to generate the topological and geometric properties of a set of rooms within a given boundary. Recently, learning-based methods have made significant progress in generating realistic floor plans. However, these methods involve a workflow of converting vector data into raster images, using image-based generative models, and then converting the results back into vector data. This process is complex and redundant, often resulting in information loss. Raster images, unlike vector data, cannot scale without losing detail and precision. To address these issues, we propose a novel deep learning framework called DiffPlanner for boundary-constrained floor plan generation, which operates entirely in vector space. Our framework is a Transformer-based conditional diffusion model that integrates an alignment mechanism in training, aligning the optimization trajectory of the model with the iterative design processes of designers. This enables our model to handle complex vector data, better fit the distribution of the predicted targets, accomplish the challenging task of floor plan layout design, and achieve user-controllable generation. We conduct quantitative comparisons, qualitative evaluations, ablation experiments, and perceptual studies to evaluate our method. Extensive experiments demonstrate that DiffPlanner surpasses existing state-of-the-art methods in generating floor plans and bubble diagrams in the creative stages, offering more controllability to users and producing higher-quality results that closely match the ground truths.

Residential design is a complex and open-ended problem that requires designers to integrate diverse types of input information while adhering to stringent energy consumption standards. However, most current research in this field focuses on generating floor plans from a limited set of input types, often neglecting to incorporate energy-related physical constraints. Existing approaches are limited by: (1) the lack of multimodal datasets in this domain, (2) the absence of comprehensive residential energy consumption data, and (3) the challenges associated with effectively integrating multiple input types into a unified model. To address these challenges, we propose MRED-14, the first large-scale Multimodal Residential Energy Dataset, comprising 14 input types, including energy consumption values, vector drawings, and textual descriptions, paired with 41,280 high-quality residential floor plans that have been scored and annotated by human experts. Based on this dataset, we introduce the LER-net model, which can flexibly adapt to various input types and generate low-energy residential floor plans. Experimental results demonstrate that LER-net outperforms existing models, achieving state-of-the-art performance under the same input conditions. In addition, the energy consumption of the generated floor plans is reduced by 5.1% compared to the actual residential designs. Further expert evaluations confirm the LER-net model's feasibility for use in residential design.

Recognition and reconstruction of residential floor plan drawings are important and challenging in design, decoration, and architectural remodeling fields. An automatic framework is provided that accurately recognizes the structure, type, and size of the room, and outputs vectorized 3D reconstruction results. Deep segmentation and detection neural networks are utilized to extract room structural information. Key points detection network and cluster analysis are utilized to calculate scales of rooms. The vectorization of room information is processed through an iterative optimization-based method. The system significantly increases accuracy and generalization ability, compared with existing methods. It outperforms other systems in floor plan segmentation and vectorization process, especially inclined wall detection.

No abstract available

No abstract available

This paper presents a smart navigation system that uses Building Information Modeling (BIM) data extracted from Industry Foundation Classes (IFC) files to generate a graph network for efficient path planning. The proposed method focuses on the construction of a graph networks, and the application of optimization algorithms to estimate optimal paths in a building environment. First, we extract relevant information about the building such as spaces, doors. This data serves to construct a graph network that represents the connectivity and relationships between different spaces in the building. Once the graph network is established, optimization algorithms are used to estimate optimal paths for navigation. The proposed method aims to provide accurate and efficient path recommendations, enhancing navigation in the building environment. The performance of our method is evaluated using a larger graph network derived from a real-world building. The results demonstrate the potential of the smart navigation system in achieving reliable path planning.

The technological innovation is touching different areas of knowlegde, including the building industry. The representation of the built environment through digital models and the inclusion of real-time information in the represented objects, assisting its operation and management, is possible using digital twins. This study aims to to adress the interoperability issues tackled by the digital models, using neural networks to integrate and generate data in an open, accessible and common language format. From a standardized test room facility model (IFC), all the different information of a real experimental procedure was embedded into python environment. Relationships were stablished according subjects’ caracteristics and identifiers, generating a new graph neural network, that associates all the relevant information to visualization and management. Deep learning algorithms supports the interpretation of larger and more complex databases, also to relationships’ prediction and classification. The proposed method assists the creation of a more integrated digital environment in the building industry.

No abstract available

No abstract available

Autonomous operation of service robotics in human-centric scenes remains challenging due to the need for understanding of changing environments and context-aware decision-making. While existing approaches like topological maps offer efficient spatial priors, they fail to model transient object relationships, whereas dense neural representations (e.g., NeRF) incur prohibitive computational costs. Inspired by the hierarchical scene representation and video scene graph generation works, we propose Hi-Dyna Graph, a hierarchical dynamic scene graph architecture that integrates persistent global layouts with localized dynamic semantics for embodied robotic autonomy. Our framework constructs a global topological graph from posed RGB-D inputs, encoding room-scale connectivity and large static objects (e.g., furniture), while environmental and egocentric cameras populate dynamic subgraphs with object position relations and human-object interaction patterns. A hybrid architecture is conducted by anchoring these subgraphs to the global topology using semantic and spatial constraints, enabling seamless updates as the environment evolves. An agent powered by large language models (LLMs) is employed to interpret the unified graph, infer latent task triggers, and generate executable instructions grounded in robotic affordances. We conduct complex experiments to demonstrate Hi-Dyna Grap's superior scene representation effectiveness. Real-world deployments validate the system's practicality with a mobile manipulator: robotics autonomously complete complex tasks with no further training or complex rewarding in a dynamic scene as cafeteria assistant. See https://anonymous.4open.science/r/Hi-Dyna-Graph-B326 for video demonstration and more details.

Abstract. Data organization is essential for effective analysis of the spatial relationships between rooms and walls. Segmentation in successive stages plays a crucial role in this process since dividing the data set into smaller sets makes its analysis easier. The proposed approach starts with the segmentation of buildings by storeys using a three-dimensional point cloud and is carried out by detecting peaks in histogram of Z frequency. Subsequently, each storey is segmented into rooms using three-dimensional mathematical morphology techniques, which allows the delimitation of the interior spaces. The third and final step consists of identifying elements within each room, such as doors, ceiling, floor, and walls. During this process, connectivity and adjacency of building elements are studied to automatically derive topological graphs. This methodology results in a deeper and more systematic analysis of three-dimensional spaces, providing a solid basis for the subsequent interpretation and manipulation of the data obtained. The proposed method has been tested in two real cases and the results are shown respectively.

The automatic generation of floorplans given user inputs has great potential in architectural design and has recently been explored in the computer vision community. However, the majority of existing methods synthesize floorplans in the format of rasterized images, which are difficult to edit or customize. In this paper, we aim to synthesize floorplans as sequences of 1-D vectors, which eases user interaction and design customization. To generate high fidelity vectorized floorplans, we propose a novel two-stage framework, including a draft stage and a multi-round refining stage. In the first stage, we encode the room connectivity graph input by users with a graph convolutional network (GCN), then apply an autoregressive transformer network to generate an initial floorplan sequence. To polish the initial design and generate more visually appealing floorplans, we further propose a novel panoptic refinement network(PRN) composed of a GCN and a transformer network. The PRN takes the initial generated sequence as input and refines the floorplan design while encouraging the correct room connectivity with our proposed geometric loss. We have conducted extensive experiments on a real-world floorplan dataset, and the results show that our method achieves state-of-the-art performance under different settings and evaluation metrics.

We present a novel and light‐weight approach to capture and reconstruct structured 3D models of multi‐room floor plans. Starting from a small set of registered panoramic images, we automatically generate a 3D layout of the rooms and of all the main objects inside. Such a 3D layout is directly suitable for use in a number of real‐world applications, such as guidance, location, routing, or content creation for security and energy management. Our novel pipeline introduces several contributions to indoor reconstruction from purely visual data. In particular, we automatically partition panoramic images in a connectivity graph, according to the visual layout of the rooms, and exploit this graph to support object recovery and rooms boundaries extraction. Moreover, we introduce a plane‐sweeping approach to jointly reason about the content of multiple images and solve the problem of object inference in a top‐down 2D domain. Finally, we combine these methods in a fully automated pipeline for creating a structured 3D model of a multi‐room floor plan and of the location and extent of clutter objects. These contribution make our pipeline able to handle cluttered scenes with complex geometry that are challenging to existing techniques. The effectiveness and performance of our approach is evaluated on both real‐world and synthetic models.

In response to the growing importance of AI-driven residential design and the lack of dedicated evaluation metrics, we propose the Residential Floor Plan Assessment (RFP-A), a comprehensive framework tailored to architectural evaluation. RFP-A consists of multiple metrics that assess key aspects of floor plans, including room count compliance, spatial connectivity, room locations, and geometric features. It incorporates both rule-based comparisons and graph-based analysis to ensure design requirements are met. A comparison of RFP-A and existing metrics was conducted both qualitatively and quantitatively, and it was revealed that RFP-A provides more robust, interpretable, and computationally efficient assessments of the accuracy and diversity of generated plans. We evaluated the performance of six existing floor plan generation models using RFP-A, showing that, surprisingly, only HouseDiffusion and FloorplanDiffusion achieved accuracies above 90%, while other models scored below or around 60%. We further conducted a quantitative comparison of diversity, revealing that FloorplanDiffusion, HouseDiffusion, and HouseGAN each demonstrated strengths in different aspects—graph structure, spatial location, and room geometry, respectively—while no model achieved consistently high diversity across all dimensions. In addition, existing metrics can not reflect the quality of generated designs well, and the diversity of the generated designs depends on both the model input and structure. Our study not only enhances the assessment of generated floor plans but also aids architects in utilizing numerous generated designs effectively.

Maintaining up-to-date environmental models from initial deployment through long-term autonomy in service is critical for applications such as navigation and task planning. To address the challenges of persistent monitoring in unknown environments, we introduce a two-stage monitoring strategy, termed the semantic-level autonomous exploration and long-term environment monitoring (SELM) framework. In the first stage, we introduce a novel semantic exploration method to adapt to new environments quickly. Leveraging the semantic information within the incrementally constructed 3-D scene graph (3-DSG), we combine the next-best-view (NBV) selection with room semantics, introducing a more efficient and comprehensive approach for multiroom indoor environment exploration. In addition, the exploration provides patrol routes, the room distance–connectivity graph, and complete environment initial states for subsequential monitoring. The monitoring stage aims to persistently patrol to update the world model in the presence of dynamic changes, including changes in objects’ positions. We formulate the long-term monitoring problem as the partially observable Markov decision process (POMDP) to cope with the environmental uncertainty. To solve the POMDP, we propose the graph attention bidirectional long short-term memory proximal policy optimization (GABPPO) algorithm for the optimal patrol strategy. The feasibility and effectiveness of the proposed SELM framework are verified through extensive experiments.

The popularity of games as an interactive entertainment medium continues to grow, with 2D maps playing a vital role in enhancing user experience. Manual map creation is time-intensive, particularly as game worlds become increasingly complex. Procedural Content Generation (PCG) offers a solution by automating map creation, improving replayability, and reducing designer workload. This research explores the use of the Binary Space Partitioning (BSP) algorithm for procedural dungeon map generation, incorporating random connections between rooms to create more exploratory and dynamic maps. The process includes three stages: developing a dungeon map generator, implementing BSP with random room connectors, and validating the generated maps to ensure navigability. Space Syntax analysis, including Visibility Graph Analysis (VGA) and Axial Line Analysis, is applied to evaluate the quality of the maps in terms of connectivity, visibility, and integration. Results show that BSP-generated maps with random connections offer dynamic layouts, while Space Syntax measures reveal that smaller minimum room sizes result in lower integration and connectivity but increase interaction hotspots. This study demonstrates the potential of BSP in generating varied game maps and the utility of Space Syntax for assessing their spatial properties.

The need for reliable internet connectivity in educational environments is crucial, but is often hampered by inefficient network infrastructure. This study aims to design an optimal Fiber To The Room (FTTR) network topology design based on Gigabit Passive Optical Network (GPON) at SMK NU Ma'arif Kudus, focusing on the efficiency of fiber optic cable installation routes. The research method used is engineering design with a quantitative approach, where the school architectural plan is modeled into a weighted graph. Route optimization is carried out by implementing the Dijkstra Algorithm to find the shortest path from the center node (ODC) to all node termination points (ODP). While node I (ODC) is designated as the starting node because it functions as the network distribution center. The calculation process is carried out by determining the minimum distance from the starting node to all destination nodes (ODP). The calculation results show that the shortest path is divided into two main routes, namely I to C to B to A to D to E and I to F to G to H. The selection of this node is proven to be able to produce a more efficient total distance compared to direct paths in several network segments. Based on these results, it can be concluded that Dijkstra's algorithm is effective in fiber optic network planning because it can optimally determine the path with the minimum distance. The application of this method is expected to assist in decision-making regarding fiber optic network infrastructure planning, making it more efficient and applicable for implementation in school environments.

Mechanical ventilation comprises a significant proportion of the total energy consumed in buildings. Sufficient natural ventilation in buildings is critical in reducing the energy consumption of mechanical ventilation while maintaining a comfortable indoor environment for occupants. In this paper, a new computerized framework based on building information modelling (BIM) and machine learning data-driven models is presented to analyze the optimum thermal comfort for indoor environments with the effect of natural ventilation. BIM provides geometrical and semantic information of the built environment, which are leveraged for setting the computational domain and boundary conditions of computational fluid dynamics (CFD) simulation. CFD modelling is conducted to obtain the flow field and temperature distribution, the results of which determine the thermal comfort index in a ventilated environment. BIM–CFD provides spatial data, boundary conditions, indoor environmental parameters, and the thermal comfort index for machine learning to construct robust data-driven models to empower the predictive analysis. In the neural network, the adjacency matrix in the field of graph theory is used to represent the spatial features (such as zone adjacency and connectivity) and incorporate the potential impact of interzonal airflow in thermal comfort analysis. The results of a case study indicate that utilizing natural ventilation can save cooling power consumption, but it may not be sufficient to fulfil all the thermal comfort criteria. The performance of natural ventilation at different seasons should be considered to identify the period when both air conditioning energy use and indoor thermal comfort are achieved. With the proposed new framework, thermal comfort prediction can be examined more efficiently to study different design options, operating scenarios, and changeover strategies between various ventilation modes, such as better spatial HVAC system designs, specific room-based real-time HVAC control, and other potential applications to maximize indoor thermal comfort.

We introduce Blueprint-Bench, a benchmark designed to evaluate spatial reasoning capabilities in AI models through the task of converting apartment photographs into accurate 2D floor plans. While the input modality (photographs) is well within the training distribution of modern multimodal models, the task of spatial reconstruction requires genuine spatial intelligence: inferring room layouts, understanding connectivity, and maintaining consistent scale. We evaluate leading language models (GPT-5, Claude 4 Opus, Gemini 2.5 Pro, Grok-4), image generation models (GPT-Image, NanoBanana), and agent systems (Codex CLI, Claude Code) on a dataset of 50 apartments with approximately 20 interior images each. Our scoring algorithm measures similarity between generated and ground-truth floor plans based on room connectivity graphs and size rankings. Results reveal a significant blind spot in current AI capabilities: most models perform at or below a random baseline, while human performance remains substantially superior. Image generation models particularly struggle with instruction following, while agent-based approaches with iterative refinement capabilities show no meaningful improvement over single-pass generation. Blueprint-Bench provides the first numerical framework for comparing spatial intelligence across different model architectures. We will continue evaluating new models as they are released and welcome community submissions, monitoring for the emergence of spatial intelligence in generalist AI systems.

We present a novel vectorized indoor modeling approach that converts point clouds into building information models (BIM) with concise and semantically segmented polygonal meshes. Existing methods detect planar shapes and connect them to complete the scene. Some focus on floor plan reconstruction as a simplified problem to better analyze connectivity between planes of floors and walls. However, the connectivity analysis is still challenging and ill-posed with incomplete point clouds as input. We propose ArrangementNet to estimate scene arrangements from an incomplete point cloud, which we can easily convert into a BIM model. ArrangementNet is a novel graph neural network that consumes noisy over-partitioned initial arrangements extracted through non-learning techniques and outputs high-quality scene arrangement. The core of ArrangementNet is an extended graph convolution that leverages co-linear and co-face relationships in the arrangement and improves the quality of prediction in complex scenes. We apply ArrangementNet to improve floor plan and ceiling arrangements and enrich them with semantic objects as scene arrangements for scene generation. Our approach faithfully models challenging scenes obtained from laser scans or multiview stereo and shows significant improvement in BIM model reconstruction compared to the state-of-the-art. Our code is available at https://github.com/zssjh/ArrangementNet.

Automated floor plan generation lies at the intersection of combinatorial search, geometric constraint satisfaction, and functional design requirements -- a confluence that has historically resisted a unified computational treatment. While recent deep learning approaches have improved the state of the art, they often struggle to capture architectural reasoning: the precedence of topological relationships over geometric instantiation, the propagation of functional constraints through adjacency networks, and the emergence of circulation patterns from local connectivity decisions. To address these fundamental challenges, this paper introduces GFLAN, a generative framework that restructures floor plan synthesis through explicit factorization into topological planning and geometric realization. Given a single exterior boundary and a front-door location, our approach departs from direct pixel-to-pixel or wall-tracing generation in favor of a principled two-stage decomposition. Stage A employs a specialized convolutional architecture with dual encoders -- separating invariant spatial context from evolving layout state -- to sequentially allocate room centroids within the building envelope via discrete probability maps over feasible placements. Stage B constructs a heterogeneous graph linking room nodes to boundary vertices, then applies a Transformer-augmented graph neural network (GNN) that jointly regresses room boundaries.

Floorplans often require considering numerous factors, from the layout size to cost, numeric attributes such as room sizes, and other intrinsic properties such as connectivity between visible regions. Representing these complex factors is challenging, but doing so in a representative and efficient way can enable new modes of design exploration. Existing image and graph-based approaches of floorplans’ representation often failed to consider low-level space semantics, structural features, and space utilization with respect to its future inhabitants, which are all the critical elements to analyze design layouts. We present a latent-space representation of floorplans using gated recurrent unit variational autoencoder (GRU-VAE), where floorplans are represented as attributed graphs (encoded with the abovementioned features). Two local search approaches are presented to efficiently explore the latent space for optimizing and generating new floorplans for the given environment. Semantic, structural, and visibility metrics are evaluated individually and as a combined objective for optimizations.

3D modelling of indoor environment is essential in smart city applications such as building information modelling (BIM), spatial location application, energy consumption estimation, and signal simulation, etc. Fast and stable reconstruction of 3D models from point clouds has already attracted considerable research interest. However, in the complex indoor environment, automated reconstruction of detailed 3D models still remains a serious challenge. To address these issues, this paper presents a novel method that couples linear structures with three-dimensional geometric surfaces to automatically reconstruct 3D models using point cloud data from mobile laser scanning. In our proposed approach, a fully automatic room segmentation is performed on the unstructured point clouds via multi-label graph cuts with semantic constraints, which can overcome the over-segmentation in the long corridor. Then, the horizontal slices of point clouds with individual room are projected onto the plane to form a binary image, which is followed by line extraction and regularization to generate floorplan lines. The 3D structured models are reconstructed by multi-label graph cuts, which is designed to combine segmented room, line and surface elements as semantic constraints. Finally, this paper proposed a novel application that 5G signal simulation based on the output structural model to aim at determining the optimal location of 5G small base station in a large-scale indoor scene for the future. Four datasets collected using handheld and backpack laser scanning systems in different locations were used to evaluate the proposed method. The results indicate our proposed methodology provides an accurate and efficient reconstruction of detailed structured models from complex indoor scenes.

ABSTRACT This article presents an approach to computing K shortest paths in large buildings with complex horizontal and vertical connectivity. The building topology is obtained from Building Information Model (BIM) and implemented using directed multigraphs. Hierarchical design allows the calculation of feasible paths without the need to load into memory the topology of the entire building. It is possible to expand the graph with new connectivity on-the-fly. The paths calculated may be composed of traversable building components that are located inside the buildings or those that are both inside and outside buildings. The performance (computational time and heap size used) is optimized by using the appropriate collections (maps, lists and sets). The proposed algorithm is evaluated in several use-case scenarios – complete graphs and real building environments. In all test scenarios, the proposed path finding algorithm is faster and uses less memory when compared to the fast version of the Yen’s KSP algorithm. The proposed approach can be successfully used as a first level of coarse-to-fine path finding algorithms.

Three-dimensional (3-D) modeling of indoor environment plays an important role in various applications such as indoor navigation, Building Information Modeling (BIM), interactive visualization, etc. While automated reconstruction of 3-D models from point clouds is receiving more and more attention. Indoor modeling remains a challenging task in terms of dealing with the complexity of indoor environment, the level of automation and restrictions of input data. To address these issues, an automatic indoor reconstruction method that quickly and effectively reconstructs indoor environment of multi-floors and multi-rooms using both point clouds and trajectories from mobile laser scanning (MLS) is proposed. The proposed automatic method of parametric structure modeling comprises three steps. First, structural elements, such as doors, windows, walls, floors, and ceilings, are extracted based on the geometric and semantic features of point clouds. Then, the point clouds are automatically segmented into adjoining rooms through a combination of visibility analysis and physical constraints of the structural elements, which ensures the integrity of the room-space partitions and yields priors for the definition of point cloud label for reconstructed model. Finally, 3-D models of individual rooms are constructed by solving an energy optimization function via multi-label graph cuts. Three benchmark datasets collected by two handheld laser scanning (HLS) and a backpack laser scanning (BLS) system were used to evaluate the proposed method. Experiments demonstrate that the recall and precision of reconstructed surface models obtained by the proposed method are mostly larger than 60%, and the average F1-score of the model is close to 5 cm.

This study is focused on indoor navigation network extraction for navigation applications based on available 3D building data and using SFCGAL library, e.g. simple features computational geometry algorithms library. In this study, special attention is given to 3D cadastre and BIM (building information modelling) datasets, which have been used as data sources for 3D geometric indoor modelling. SFCGAL 3D functions are used for the extraction of an indoor network, which has been modelled in the form of indoor connectivity graphs based on 3D geometries of indoor features. The extraction is performed by the integration of extract transform load (ETL) software and the spatial database to support multiple data sources and provide access to SFCGAL functions. With this integrated approach, the current lack of straightforward software support for complex 3D spatial analyses is addressed. Based on the developed methodology, we perform and discuss the extraction of an indoor navigation network from 3D cadastral and BIM data. The efficiency and performance of the network analyses were evaluated using the processing and query execution times. The results show that the proposed methodology for geometry-based navigation network extraction of buildings is efficient and can be used with various types of 3D geometric indoor data.

No abstract available

The paper presents a novel approach for vector-floorplan generation via a diffusion model, which denoises 2D coordinates of room/door corners with two inference objectives: 1) a single-step noise as the continuous quantity to precisely invert the continuous forward process; and 2) the final 2D coordinate as the discrete quantity to establish geometric incident relationships such as parallelism, orthogonality, and corner-sharing. Our task is graph-conditioned floorplan generation, a common workflow in floorplan design. We represent a floorplan as 1D polygonal loops, each of which corresponds to a room or a door. Our diffusion model employs a Transformer architecture at the core, which controls the attention masks based on the input graph-constraint and directly generates vector-graphics floorplans via a discrete and continuous denoising process. We have evaluated our approach on RPLAN dataset. The proposed approach makes significant improvements in all the metrics against the state-of-the-art with significant margins, while being capable of generating non-Manhattan structures and controlling the exact number of corners per room. A project website with supplementary video and document is here https://aminshabani.github.io/housediffusion.

Floorplanning, as an early step in physical design, will greatly affect the PPA of the later stages. To achieve better performance while main-taining relatively the same chip size, the utilization of the generated floorplan needs to be high and constraints related to design rules, routability, power should be honored. In this paper, we propose a two-step framework, called TOFU, for floorplan whitespace reduction with fixed-outline and soft/pre- placed/hard modules modeled. Whitespace is first reduced by iteratively refining the locations of modules. Then the modules near whitespace will be changed into rectilinear shapes to further improve the utilization. To ensure the legality and quality of the intermediate floorplan during the refinement process, a constraint graph-based legalizer with a novel constraint graph construction method is proposed. Experimental results show that the whitespace of the initial floorplans generated by Corblivar [1] can be reduced by about 70% on average and up to 90% in several cases. Moreover, the resulting wirelength is also 3% shorter due to a higher utilization.

This paper proposes a novel graph-constrained generative adversarial network, whose generator and discriminator are built upon relational architecture. The main idea is to encode the constraint into the graph structure of its relational networks. We have demonstrated the proposed architecture for a new house layout generation problem, whose task is to take an architectural constraint as a graph (i.e., the number and types of rooms with their spatial adjacency) and produce a set of axis-aligned bounding boxes of rooms. We measure the quality of generated house layouts with the three metrics: the realism, the diversity, and the compatibility with the input graph constraint. Our qualitative and quantitative evaluations over 117,000 real floorplan images demonstrate that the proposed approach outperforms existing methods and baselines. We will publicly share all our code and data.

ACG (adjacent constraint graph) is invented as a general floorplan representation. It has advantages of both adjacency graph and constraint graph of a floorplan: edges in an ACG are between modules close to each other, thus the physical distance of two modules can be measured directly in the graph; since an ACG is a constraint graph, the floorplan area and module positions can be simply found by longest path computations. A natural combination of horizontal and vertical relations within one graph renders a beautiful data structure with full symmetry. The direct correspondence between geometrical positions of modules and ACG structures also makes it easy to incrementally change a floorplan and evaluate the result. Experimental results show the superiority of this representation.

Floorplanning has long been a critical physical design task with high computation complexity. Its key objective is to determine the initial locations of macros and standard cells with optimized wirelength for a given area constraint. This paper presents Flora, a graph attentionbased floorplanner to learn an optimized mapping between circuit connectivity and physical wirelength, and produce a chip floorplan using efficient model inference. Flora has been integrated with two state-of-the-art mixed-size placers. Experimental studies using both academic benchmarks and industrial designs demonstrate that compared to state-of-the-art mixed-size placers alone, Flora improves placement runtime by 18%, with 2% wirelength reduction on average.

Chip floorplanning has long been a critical task with high computation complexity in the physical implementation of VLSI chips. Its key objective is to determine the initial locations of large chip modules with minimized wirelength while adhering to the density constraint, which in essence is a process of constructing an optimized mapping from circuit connectivity to physical locations. Proven to be an NP-hard problem, chip floorplanning is difficult to be solved efficiently using algorithmic approaches. This article presents GraphPlanner, a variational graph-convolutional-network-based deep learning technique for chip floorplanning. GraphPlanner is able to learn an optimized and generalized mapping between circuit connectivity and physical wirelength and produce a chip floorplan using efficient model inference. GraphPlanner is further equipped with an efficient clustering method, a unification of hyperedge coarsening with graph spectral clustering, to partition a large-scale netlist into high-quality clusters with minimized inter-cluster weighted connectivity. GraphPlanner has been integrated with two state-of-the-art mixed-size placers. Experimental studies using both academic benchmarks and industrial designs demonstrate that compared to state-of-the-art mixed-size placers alone, GraphPlanner improves placement runtime by 25% with 4% wirelength reduction on average.

No abstract available

No abstract available

No abstract available

No abstract available

No abstract available

This paper proposes a generative adversarial layout refinement network for automated floorplan generation. Our architecture is an integration of a graph-constrained relational GAN and a conditional GAN, where a previously generated layout becomes the next input constraint, enabling iterative refinement. A surprising discovery of our research is that a simple non-iterative training process, dubbed component-wise GT-conditioning, is effective in learning such a generator. The iterative generator further allows us to improve a metric of choice via meta-optimization techniques by controlling when to pass which input constraints during iterative refinement. Our qualitative and quantitative evaluation based on the three standard metrics demonstrate that the proposed system makes significant improvements over the current state-of-the-art, even competitive against the ground-truth floorplans, designed by professional architects. Code, model, and data are available at https://ennauata.github.io/houseganpp/page.html.

This paper addresses the challenge of object-centric lay-out generation under spatial constraints, seen in multi-ple domains including floorplan design process. The de-sign process typically involves specifying a set of spa-tial constraints that include object attributes like size and inter-object relations such as relative positioning. Existing works, which typically represent objects as single nodes, lack the granularity to accurately model complex interactions between objects. For instance, often only certain parts of an object, like a room's right wall, interact with adjacent objects. To address this gap, we introduce a factor graph based approach with four latent variable nodes for each room, and a factor node for each constraint. The factor nodes represent dependencies among the variables to which they are connected, effectively capturing constraints that are potentially of a higher order. We then develop message-passing on the bipartite graph, forming a factor graph neu-ral network that is trained to produce a floorplan that aligns with the desired requirements. Our approach is simple and generates layouts faithful to the user requirements, demon-strated by a large improvement in IOU scores over existing methods. Additionally, our approach, being inferential and accurate, is well-suited to the practical human-in-the-loop design process where specifications evolve iteratively, offering a practical and powerful tool for AI-guided design.

No abstract available

No abstract available

This paper proposes a novel generative adversarial layout refinement network for automated floorplan generation. Our architecture is an integration of a graph-constrained relational GAN and a conditional GAN, where a previously generated layout becomes the next input constraint, enabling iterative refinement. A surprising discovery of our research is that a simple non-iterative training process, dubbed component-wise GT-conditioning, is effective in learning such a generator. The iterative generator also creates a new opportunity in further improving a metric of choice via meta-optimization techniques by controlling when to pass which input constraints during iterative layout refinement. Our qualitative and quantitative evaluation based on the three standard metrics demonstrate that the proposed system makes significant improvements over the current state-of-the-art, even competitive against the ground-truth floorplans, designed by professional architects.

No abstract available

No abstract available

In this paper we present a timing-influenced floorplanner for general cell IC design. The floorplanner works in two phases. In the first phase we restrict the modules to be rigid and the floorplan to be slicing. The second phase of floorplanner allows modification to the aspect ratios of individual modules to further reduce the area of the overall bounding box. The first phase is implemented using genetic algorithm while in the second phase we adopt a constraint graph based approach. Experimental results are also presented.

ACG (adjacent constraint graph) is a general floorplan representation. The refinement of constraint graphs gives not only an efficient representation but also a representation sharing the advantage of adjacency graphs. As most edges in an ACG are between modules that are close to each other, the physical-distance of two modules can be measured without packing by the shortest path between them on the ACG. Experimental results verified this relationship and possible approaches for interconnect planning are discussed.

No abstract available

No abstract available

No abstract available

No abstract available

No abstract available

No abstract available

No abstract available

Automated floorplanning or space layout planning has been a long-standing NP-hard problem in the field of computer-aided design, with applications in integrated circuits, architecture, urbanism, and operational research. In this paper, we introduce GenFloor, an interactive design system that takes geometrical, topological, and performance goals and constraints as input and provides optimized spatial design solutions as output. As part of our work, we propose three novel permutation methods for existing space layout graph representations, namely O-Tree and B*-Tree representations. We implement our proposed floorplanning methods as a package for Dynamo, a visual programming tool, with a custom GUI and additional evaluation functionalities to facilitate designers in their generative design workflow. Furthermore, we illustrate the performance of GenFloor in two sets of case-study experiments for residential floorplanning tasks by (a) measuring the ability of the proposed system to find a known optimal solution, and (b) observing how the system can generate diverse floorplans while addressing given a constant residential design problem. Our results indicate convergence to the global optimum is achieved while offering a diverse set of solutions of a residential floorplan corresponding to the Pareto-optimums of the solution landscape.

. In many aspects of architectural design, the optimization of spatial layout is particularly important, as it is directly related to the practicality, comfort, and aesthetics of the building. The objective of this research is to investigate the utilization of graph neural networks (GNN) and computer-aided design (CAD) in refining the layout of architectural spaces. In response to the disorderly expansion in the current building configuration, the development of urbanization has entered a relatively complex relationship. This article introduces a visual quick layout framework for building configurations. Analyzing the inherently complex relationship between GNN and architectural space constructs a visual architectural spatial configuration. Compared with traditional methods, the overall design appears more coherent and efficient. Among different functional requirements, it has higher functional requirements in method layout, which improves the efficiency of architectural design in overall design. In improving the architectural layout of GNN and CAD, the development of urban planning is also reflected differently in the integration of regions.

This research presents an intelligent optimization algorithm framework for chain restaurant spatial layout generation based on Generative Adversarial Networks (GANs). Contemporary restaurant design methodologies rely on subjective expertise and static planning approaches that inadequately address dynamic operational requirements and evolving consumer preferences. The proposed GAN-based architecture incorporates a dual-generator framework with progressive upsampling modules and multi-head attention mechanisms specifically designed for restaurant spatial optimization. The multi-objective optimization function integrates operational efficiency metrics, spatial utilization coefficients, and aesthetic quality assessments through weighted objective aggregation, achieving balanced performance across competing design criteria. Experimental validation utilizing 3,892 restaurant layouts across 47 chain brands demonstrates substantial improvements in spatial layout quality metrics. Generated layouts achieve average efficiency scores of 87.3% compared to traditional baseline measurements of 72.8%, representing a 19.9% performance enhancement. The algorithm reduces average customer movement distances by 23.4% while maintaining 92.6% regulatory compliance rates. Implementation case studies across three distinct restaurant chains validate practical deployment feasibility with measurable improvements in operational efficiency ranging from 12.9% to 24.6%. The research establishes foundational technologies for next-generation intelligent restaurant design systems that enable data-driven optimization while reducing traditional design development timelines by approximately 65%. Commercial deployment analysis indicates potential cost savings of $12,000-$18,000 per location through reduced architectural consultation requirements.

Recent advancements in AI research, particularly in spatial layout generation, highlight its capacity to enhance human creativity by swiftly providing architects with numerous alternatives during the pre-design phase. The complexity of architectural design data, characterized by multifaceted elements and varying representations, presents significant challenges in creating uniform and robust datasets. This study addresses this challenge by developing a robust training dataset specifically tailored for AI-driven spatial layout generation in architecture. An algorithm capable of extracting spatial relationship diagrams from raster-based floor plan images and converting them into vector-based data was introduced. Through extensive web crawling, a dataset comprising 10,000 data rows, categorized into 21 classes and three spatial relationship categories, was collected. When tested with the You-Only-Look-Once (YOLO) model, the detection rate was 99%, the mean average precision was 85%, and the MIoU was 74.2%. The development of this robust training dataset holds significant potential to advance knowledge-based artificial intelligence design automation studies, paving the way for further innovation in architectural design.

We present a novel graph Transformer generative adversarial network (GTGAN) to learn effective graph node relations in an end-to-end fashion for challenging graph-constrained architectural layout generation tasks. The proposed graph-Transformer-based generator includes a novel graph Transformer encoder that combines graph convolutions and self-attentions in a Transformer to model both local and global interactions across connected and non-connected graph nodes. Specifically, the proposed connected node attention (CNA) and non-connected node attention (NNA) aim to capture the global relations across connected nodes and non-connected nodes in the input graph, respectively. The proposed graph modeling block (GMB) aims to exploit local vertex interactions based on a house layout topology. Moreover, we propose a new node classification-based discriminator to preserve the high-level semantic and discriminative node features for different house components. To maintain the relative spatial relationships between ground truth and predicted graphs, we also propose a novel graph-based cycle-consistency loss. Finally, we propose a novel self-guided pre-training method for graph representation learning. This approach involves simultaneous masking of nodes and edges at an elevated mask ratio (i.e., 40%) and their subsequent reconstruction using an asymmetric graph-centric autoencoder architecture. This method markedly improves the model's learning proficiency and expediency. Experiments on three challenging graph-constrained architectural layout generation tasks (i.e., house layout generation, house roof generation, and building layout generation) with three public datasets demonstrate the effectiveness of the proposed method in terms of objective quantitative scores and subjective visual realism. New state-of-the-art results are established by large margins on these three tasks.

Visual simultaneous localization and mapping (SLAM) systems face challenges in detecting loop closure under the circumstance of large viewpoint changes. In this paper, we present an object-based loop closure detection method based on the spatial layout and semanic consistency of the 3D scene graph. Firstly, we propose an object-level data association approach based on the semantic information from semantic labels, intersection over union (IoU), object color, and object embedding. Subsequently, multi-view bundle adjustment with the associated objects is utilized to jointly optimize the poses of objects and cameras. We represent the refined objects as a 3D spatial graph with semantics and topology. Then, we propose a graph matching approach to select correspondence objects based on the structure layout and semantic property similarity of vertices' neighbors. Finally, we jointly optimize camera trajectories and object poses in an object-level pose graph optimization, which results in a globally consistent map. Experimental results demonstrate that our proposed data association approach can construct more accurate 3D semantic maps, and our loop closure method is more robust than point-based and object-based methods in circumstances with large viewpoint changes.

ABSTRACT Building façades can feature different patterns depending on the architectural style, functionality, and size of the buildings; therefore, reconstructing these façades can be complicated. In particular, when semantic façades are reconstructed from point cloud data, uneven point density and noise make it difficult to accurately determine the façade structure. When investigating façade layouts, Gestalt principles can be applied to cluster visually similar floors and façade elements, allowing for a more intuitive interpretation of façade structures. We propose a novel model for describing façade structures, namely the layout graph model, which involves a compound graph with two structure levels. In the proposed model, similar façade elements such as windows are first grouped into clusters. A down-layout graph is then formed using this cluster as a node and by combining intra- and inter-cluster spacings as the edges. Second, a top-layout graph is formed by clustering similar floors. By extracting relevant parameters from this model, we transform semantic façade reconstruction to an optimization strategy using simulated annealing coupled with Gibbs sampling. Multiple façade point cloud data with different features were selected from three datasets to verify the effectiveness of this method. The experimental results show that the proposed method achieves an average accuracy of 86.35%. Owing to its flexibility, the proposed layout graph model can deal with different types of façades and qualities of point cloud data, enabling a more robust and accurate reconstruction of façade models.