Search

The proliferation of indoor maps is limited by the manual process of generalizing floor plans. Previous attempts at automating similar processes use rasterization. With Graph Neural Networks (GNN) it is now possible to rely on the inherentstructures in CAD drawings and thus avoid rasterization. Localization of doors is a core component in floor plan generalization. We show how state-of-the-art GNNs can be used to classify graph nodes, extracted from CAD primitives, as dooror non-door. Generalization is represented by the creation of bounding boxes from the labelled graph nodes. Our baseline is a standard raster-based bounding box detector. To support further development of graph-based methods and comparison with raster-based methods, we publish a new dataset that consists of both image and graph-based floor plan representations. Our graph-based approach completely outperforms the Faster-RCNN baseline, which fail to locate any doors with thedesired localization accuracy. Code and dataset is available at https://github.com/Chrps/MapGeneralization.

Wireless communication and sensor networks have been widely employed in our daily life. Sensing-range characteristics, inter-nodal connectivity, and communication capacity are critical factors in providing reliable quality-of-service. Placement of sensors or access-points indoors is always critical to the effectiveness of a wireless communication/sensor network. In this paper, we would like to investigate the important classical sensor/access-point placement problem in the discrete-graph domain. Instead of the conventional trial-and-error approach upon rigorous quality-of-service measurements, we study the sensor/access-point coverage-area maximization problem over visibility graphs and then design a new graph-theoretical approach to deal with it. This new approach just takes the bird's eye view of an indoor geometry (floor plan) as the only necessary input information and does not rely on any quality-of-service measurement. Thus, the crucial coverage-area maximization problem for sensor/access-point placement can be tackled algorithmically. Our new approach can be applied for indoor geometries as arbitrary polygons, either convex or nonconvex.

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.

Building Information Modelling (BIM) is a process that contain all the necessary information to manage the construction project across all its lifecycle. This benefits not only the construction industry but other industries such as utility companies that need to perform tasks inside of buildings and will need to access information about its elements. However, one of the biggest challenges is the digitalisation of the existing infrastructure. The use of semantic segmentation techniques could enable the transformation of infrastructure legacy data, such as 2D floor plans images, to open-standard BIM models. In this paper, we propose a processing pipeline to transform 2D floor plan images into BIM models. The pipeline makes use of an interval Type-2 Fuzzy Rule-based System (FRBS) that has an Intersection over Union metric value of 98.62% outperforming the Type-1 version of the model. Moreover, the proposed model is highly transparent, and it allows end-users to augment it using expert knowledge, something that is not possible in deep learning opaque-box models.

We present Zillow Indoor Dataset (ZInD): A large indoor dataset with 71,474 panoramas from 1,524 real unfurnished homes. ZInD provides annotations of 3D room layouts, 2D and 3D floor plans, panorama location in the floor plan, and locations of windows and doors. The ground truth construction took over 1,500 hours of annotation work. To the best of our knowledge, ZInD is the largest real dataset with layout annotations. A unique property is the room layout data, which follows a real world distribution (cuboid, more general Manhattan, and non-Manhattan layouts) as opposed to the mostly cuboid or Manhattan layouts in current publicly available datasets. Also, the scale and annotations provided are valuable for effective research related to room layout and floor plan analysis. To demonstrate ZInD’s benefits, we benchmark on room layout estimation from single panoramas and multi-view registration.

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.

Solving robotic navigation tasks via reinforcement learning (RL) is challenging due to their sparse reward and long decision horizon nature. However, in many navigation tasks, high-level (HL) task representations, like a rough floor plan, are available. Previous work has demonstrated efficient learning by hierarchal approaches consisting of path planning in the HL representation and using sub-goals derived from the plan to guide the RL policy in the source task. However, these approaches usually neglect the complex dynamics and sub-optimal sub-goal-reaching capabilities of the robot during planning. This work overcomes these limitations by proposing a novel hierarchical framework that utilizes a trainable planning policy for the HL representation. Thereby robot capabilities and environment conditions can be learned utilizing collected rollout data. We specifically introduce a planning policy based on value iteration with a learned transition model (VI-RL). In simulated robotic navigation tasks, VI-RL results in consistent strong improvement over vanilla RL, is on par with vanilla hierarchal RL on single layouts but more broadly applicable to multiple layouts, and is on par with trainable HL path planning baselines except for a parking task with difficult non-holonomic dynamics where it shows marked improvements., Comment: 7 pages, 5 figures, 2021 IEEE International Conference on Robotics and Automation (ICRA), v2: DOI number added

The Internet of Things based Smart Sensing and Alarm System with Autonomous Guiding Robots for Efficient Fire Emergency Evacuation System was designed and developed to provide immediate monitoring and alerts, allowing for efficient emergency evacuation with robotic assistance. The project includes two types of robots: a robot guide and a robot transporter. The NodeMCU-controlled Vex Navigating Robot directs people to the closest and most appropriate safety exits, while the Remote-Controlled Arduino-based Robot Transporter shuttles restrained or elderly people. The system employs an Internet of Things-based microcontroller to read and process data from sensors that detect temperature, humidity, and the number of people present in each room. This data is sent to the Cloud server and shown in the Blynk Mobile App. When a fire occurs, an alert signal using GSM and GPS technologies is sent to the fire station, the building health and safety team, and the autonomous guiding robot that replaces the Fire Warden, informing them of the incident's location and possible casualties. In this system, the robot system plays an important role in the evacuation process, assisting in the avoidance of delays and the saving of lives. The Autonomous navigating Robot was tested in a Floor plan layout with 9 rooms. The sensing unit was fixed in all rooms and the Robot in front of BS109 is tested with 3 different sources of fire, one in BS108, BS103 and BS 110. The robot successfully chose the appropriate exit path of path2 when the source of fire is in BS103 and BS110 and path1 when source of Fire is BS108. Testing of the algorithm is done with robot in front of Bs104, BS109, BS110 with 3 different source of fire. Transporting Robot is tested with remote controlled unit and is found to successfully transport people up to the weight of 80kgs with less than 2% deviation from its destined angle of travel.

At present, fingerprinting based localization methods gain a massive attraction for numerous indoor location-based services. However, creation of a radio map is a laborious and time-consuming process that limits the applicability of this method. In this study, we propose FaStER, a Fast, Stable, Expendable and Reliable radio map creation system for fingerprinting based indoor localization applications. Data from a chest mounted IMU with ZigBee protocol based wireless network technology and indoor floor plan are used as inputs to the system. The proposed system ensures high accuracy both in terms of location estimation and RSSI acquisition by correcting errors generated from the IMU and from the presence of human body. The stability of a generated radio map is guaranteed by exploiting the concept of RSSI similarity index. The system is especially proposed for facilitating indoor location-based services with chest mounted wearable devices towards IoT connected wearable health and environmental monitoring applications. Results demonstrate that the proposed system can create a fast, stable, expendable, and reliable radio map that can help to improve the accuracy for location fingerprinting.

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.

Trajectory analysis can be useful to understand pedestrian behaviour in indoor environments. Collecting real data however can be difficult and time consuming, and many publicly available data sets and trajectory generation methods focus on outdoor scenarios. We propose a method for generating realistic indoor trajectory data, taking into account the various restrictions of indoor environments. The result of our experiments, conducted on two museum floor plans, is a system which only requires a set of points of interest and a floor plan picture. A graph is automatically generated using Delaunay triangulation and trajectories can be obtained by simulating a person walking through the graph.

In ultra-deep sub-micron technology, Routing has become challenging due to ever increasing number of metal layers, distinct layer thicknesses, new design rules and design complexity. Due to increase in congestion with lower technology nodes designer has to predict during floor plan weather routing is possible with meeting timing requirements. It is not enough to route only but need to route with DRC clean and without degrading post layout timings. Area reduction with lower node technology, routing congestion increases on chip and in that scenario need to route the best topology with metal layers. This article will discuss about routing a net with different routing topologies and the best topology which meets timing and DRC requirements will be routed. Flow is developed which will select the best topology from available topologies on the basis of different criteria and net will be routed automatically with given metal layers. Once it is routed, It will also fix the timing violations on the net with different layout solutions.

This paper provides supply-induced jitter-aware sensitivity and systematic modeling approach for analyzing supply noise induced timing jitter in multiple IPs (intellectual property) high speed interface circuits integrated in one 1-2-14L FCCSP package substrate to perform the jitter generated by the supply noise in the time domain compared to present general analysis methods. During product design phase, IP integration effort takes place very first to define the package floor plan and then this floor structure will be used for mother board design. We propose an efficient method during IP integration for verification plan for timing budget calculation when IPs work simultaneously in a single package substrate at early design stage. The goal is to provide adequate performance for cost-effective and system solution and achieving on system-level success.

This study proposes a rent-prediction model that allows users to understand which parts of a floor-plan image (FPI) significantly affect a rent prediction. Principal Component Analysis (PCA) and Bag of Features (BoF) are used to extract feature vectors from an FPI. The importance of each feature vector, derived from the FPI, is measured and visualized based on 100,000 rental properties sampled from the LIFULL HOME'S dataset. The experimental result suggests that the wall and orientation symbol drawn on an FPI contribute to the rent prediction for both PCA and BoF, although BoF provides a better understanding of which parts of an FPI significantly affect the rent prediction.

Smartphones currently support the augmented reality feature, capable of displaying virtual objects into the real world environment at a specific location designated with a marker. This research develops an augmented reality application on a smartphone to visualize a 3D house design. The floorplan image captured by the smartphone camera is designated as a marker and then sent to the server. The server-side application, built using deep learning and integer programming techniques, detects corner positions on the image and produces a vector containing 2D coordinates. The client application on the smartphone uses the 2D coordinates to draw a 3D house model using Unity 3D. The model is then added over the floor plan image marker using Vuforia. Based on the test results, the 3D house models can be drawn according to the provided floorplan with moderate accuracy. The augmented reality application offers an alternative approach to examine a floorplan so that a user can better understand and engage with the design.

The ability to search for objects is a precondition for various robotic tasks. In this paper, we address the problem of finding objects in partially known indoor environments. Using the knowledge of the floor plan and the mapped objects, we consider object-object and object-room co-occurrences as prior information for identifying promising locations where an unmapped object can be present. We propose an efficient search strategy that determines the best pose of the robot based on the analysis of the candidate locations. We optimize the probability of finding the target object and the distance travelled through a cost function.To evaluate our method, several experiments in simulated and real-world environments were performed. The results show that the robot successfully finds the target object in the environment while covering only a small portion of the search space. The real-world experiments with the TurtleBot 2 mobile robot validate the proposed approach and demonstrate that the method performs well also in real environments.

Many applications with mobile robots require self-localization in indoor maps. While such maps can be previously generated by SLAM strategies, there are various localization approaches that use 2D floor plans as reference input. In this paper, we present a localization strategy using floor plan as map, which is based on spatial density information computed from dense depth data of RGB-D cameras. We propose an interval-based model, called Interval Free-Space Density, that bounds the uncertainty of observations and minimizes the effects of movable objects in the environment. Our model was applied in a Monte Carlo Localization strategy and compared with traditional observation models. The results of experiments showed the robustness of the proposed method in single-camera and multi-camera experiments in home environments.

We propose an accessible indoor navigation application. The solution integrates information of floor plans, Bluetooth beacons, Wi-Fi/cellular data connectivity, 2D/3D visual models, and user preferences. Hybrid models of interiors are created in a modeling stage with Wi-/cellular data connectivity, beacon signal strength, and a 3D spatial model. This data is collected, as the modeler walks through the building, and is mapped to the floor plan. Client-server architecture allows scaling to large areas by lazy-loading models according to beacon signals and/or adjacent region proximity. During the navigation stage, a user with the designed mobile app is localized within the floor plan, using visual, connectivity, and user preference data, along an optimal route to their destination. User interfaces for both modeling and navigation use visual, audio, and haptic feedback for targeted users. While the current pandemic event precludes our user study, we describe its design and preliminary results.

Inventory management is the key to an efficient supply chain. Numerous kinds of inventories are stored in a warehouse where locating goods can be troublesome for the employees. Specifically, in corrugated box company, inventories of paper rolls are placed in a huge warehouse where there are hundreds of paper rolls stock ranging from different weights and sizes. Finished goods are also stored in a warehouse awaiting delivery, placed in between hundreds of other customers' orders where delivery schedule varies from one another. Thus, the positioning of the goods is not permanent and is changing along with the incoming orders. The research aims to explore BLE beacons for the most effective solution in handling inventory placement problems. Common tracking technologies for inventories include barcodes, QR codes, and RFID. The fourth industrial revolution drives the application of IoT, Bluetooth low energy (BLE) beacons are one of the examples in the area. It is the successor of the previous Bluetooth technology, which is more eminent in energy saving. In warehouses, BLE beacons are positioned to determine zone mapping in organizing the floor plan, assists in indoor navigation for guiding employees, and support asset tracking with beacon tags for identification and goods tracing.

Position estimation in localization and tracking is often approached with Bayesian filtering. In combination with Gaussian assumptions about state distributions this approach results in efficient algorithms such as the Kalman filter. Indoors, however, the location state space is naturally constrained by walls and other obstacles that cannot be taken into account with Kalman or other Gaussian filters. More general Hidden Markov Models (HMMs) on the other hand do not make restrictive assumptions and thus allow for extensive explicit modelling. In particular for indoor positioning, we show how to inject geometric apriori knowledge, i.e. a given floor plan, into the parameters of a discrete-state HMM. In contrast to classical position estimation, we use an adapted Viterbi-algorithm for decoding the maximum aposteriori (MAP) trajectory estimate based on the geometry-aware HMM. Results and advantages of the presented method are discussed for the localization of a mobile sensor node in a simulated measurement setup.

With the continuous development of computer technology, both hardware and software equipment have undergone upgrading. At the same time, in order to meet the aesthetic needs of the public and the application needs of various industries, the computer is directly used in graphic design, which can give graphic design Bring new changes. Designers have gradually developed from the use of paper and pen as a carrier to use a variety of design software as a carrier to make the designed floor plan more in line with the aesthetic needs of modern people. From the design concept to the presentation of works, continuous innovation can be achieved, thereby expanding the scope of graphic design applications. This article is based on the use of Flash AIR programming and development + Photoshop technology to build a flat visual design platform to meet the needs of customers in the graphic design process. It has the aesthetics of graphic layout, shortens creative design time, improves work efficiency, safe and reliable system, and convenient operation.

VLSI Floor-planning is known to be a NP hard problem. Therefore, most of the EDA tools use heuristic algorithms to solve this problem that yield near optimal solutions. Simulated annealing has been known to produce good results in floor planning. However, the results of simulated annealing can be biased by the initial placement of the modules. In this paper we study the effect of initial placement on the floor planning. The method can be used on slicing and non- slicing floor-plans. The proposed algorithm uses the number of blocks as specified in Microelectronics Centre of North Carolina (MCNC) benchmark circuits. The results are obtained for different configurations of a single floor plan and the best floor-plan is chosen as a near optimal solution.

A novel "physics-free" approach of designing indoor radio dot layout for a floor plan is introduced by formulating it as an image-to-image translation problem and solved with customized dimension-aware conditional generative adversarial networks (DA-cGANs). The proposed model generates a desirable radio heatmap and its respective radio dot layout from a given floor plan with wall types, physical dimension, and macro-cell interference, by learning from the accumulated indoor radio designs by human experts. Considering the nature of radio propagation, two new loss functions and a two-stage training strategy are proposed for the generator to learn the right direction of signal propagation and precise dot locations, in addition to a sectional analysis for dealing with large floor plans. Experimental results show that the new model is effectively generating acceptable dot layout designs and that dimension-awareness is a key enabler for this type of prediction.

The purpose of this study is to explore how the artificial intelligence system of floor plan assist novice designer to improve problem-solving ability of interior floor plan. Through two experiments of drawing floor plan before and after, we used the method of document analysis to compare the differences of two drawings. Novices drew a floor plan of bedroom in first stage, and next, novices operated the artificial system of interior floor plan to construct their knowledge and learned various possible schemes in a short period of time. Finally, novices drew second floor plan of bedroom again. The result shows that the second floor plan to be improved significantly after novices used the intelligence system of floor plan.

We describe and experimentally demonstrate a practical method for indoor localization using measurements obtained from resource-constrained devices with limited sensing capabilities. We focus on handheld/mobile devices but the method can be useful for a variety of wearable devices. Our system works with sparse WiFi or image-based measurements, avoiding laborious site surveying for dense signal maps and runs in real-time. It uses Conditional Random Fields to infer the most probable sequence of agent positions from a known floor plan, dead reckoning and sparse absolute position estimates. Our solution leverages known topology of the environment by pre-computing allowed motion sequences of an agent, which are then used to constraint the motion inferred from the sensory data. The system is evaluated in a typical office building, demonstrating good accuracy and robustness to sparse, low-quality measurements.

Since vertical planes and lines are rich and unique for indoor localization, such as walls, posters, frames of doors and windows, a lightweight hybrid modeling method based on simple spatial features is proposed, specifically for the indoor localization and navigation applications. Firstly, the vertical planes in 3D space are detected and projected to generate a simplified floor plan of indoor environment. Here, a strategy of region segmentation and georeferencing is employed. Moreover, the plane-based model is further optimized by iterative merging algorithm. Then, some remarkable vertical lines are detected and localized based on the plane-based model, which can compensate the inaccuracy of plane ranges and the deficiency of remarkable planes in some areas. The dynamic time warping (DTW) algorithm is applied to match homonymous lines between sequential image frames, in cooperation with Kalman filtering for simultaneous tracking of vertical lines. Finally, a hybrid indoor model consisting of vertical planes and lines is generated, which is much more lightweight than previous 3D modeling methods. Experiments in real indoor scenes using single off-the-shelf smartphone verify the feasibility and effectiveness of the proposed method. The whole model size of an indoor environment with more than 200m2 area is only 302 Bytes, and it achieves an accuracy of about 0.3 meters.

Autonomous navigation is of great importance for service robots. Such robots need various technologies, especially localization, and mapping. In this paper, we focus on the localization problem. Commonly, to solve the localization problem, the creation of a map is needed. However, creating the map takes considerable time and costs. Therefore, one solution for indoor environment is to use architectural floor plans since buildings have own floor plans. If a robot can use them for localization, it enables users to cut the time and costs for creating the map from scratch. However, the floor plans sometimes do not match with a real building. Besides, sensor measurement sometimes contains objects such as a pedestrian, which are not contained in the floor plans. In this paper, we propose a localization algorithm with architectural floor plans that is robust to such inconsistencies. The algorithm estimates a robot coordinate by matching the floor plan with the point clouds obtained from depth images. Outliers derived from the inconsistencies in the point clouds are filtered with plane information from the depth images with the Generalized ICP framework. We tested our algorithm with floor plans in a real building and in a simulator as a case study. The results show that our algorithm can localize a robot with average more than twice accuracy compared to AMCL and be used for real-time applications.

Laser scanners are becoming more affordable. The cost of a laser scanner has been driven down by the high demand for range measuring devices used in autonomous cars. The objective of this paper is to evaluate usage of a cost effective laser scanner for rapid mapping of a building’s interior. One of the valuable applications of such devices is for quick structural assessments during reconnaissance efforts right after an earthquake, when measuring a plan of the interior of the building is beneficial for further analysis. Due to safety and the large number of buildings under investigation, researchers have a very limited amount of time to remain inside of the building and get the critical measurements. Having a tool that can create layouts of the interior while a person is moving inside would be beneficial for applications of this kind. For mass application, cost of the device is an important factor. The scanner used in this application creates a plane point cloud. In a horizontal orientation, it creates a horizontal slice of the building’s interior that can capture the locations of the openings (doors and windows) and the general dimensions of the room. When the stitching of the laser scans is done on the fly, the floor plan of a large building can be measured with adequate accuracy in a short period of time. The point clouds registered in real-time and transferred over a network have a strong potential to become a very important component in the Internet of Things, a system of interrelated computing devices, mechanical and digital machines, and sensors. This work assumes that information about the orientation of the scanner and its relative location is very limited and as such, the registration of the point clouds is based on the use of common features.

In this paper, we describe a method for global localization in a previously unvisited environment using only a schematic floor plan as a prior map. The floor plan need not be a precision map–it can be the sort of image found in buildings to guide people or aid evacuation. The core idea is to identify features that are stable across both a drawn floor plan and robot point-of-view LIDAR data, for example wall intersections, which appear as corners from overhead and as vertical lines from the ground. We introduce a factor graph-based global localization method that uses these features as landmarks. The detections of such descriptorless features are noisy and often ambiguous. We therefore propose robust data association based on a pairwise measurement consistency check and maxmixtures error model. We evaluate the resulting system in a real-world indoor environment, demonstrating performance comparable to a baseline system that uses a conventional LIDAR-based prior map.

“UCSI Campus Navigator” was initiated due to real life situation of students who have problems navigating routes. The campus navigator application is an android mobile application used for navigating routes inside Block C of UCSI University. This is to provide a navigator application for students to navigate the direction and allow the students to check the classroom names based on the map of floor plan. The app will also suggest the right route to the user to ease the students to find the classroom. Data collection was done through questionnaires and observation of UCSI students. The “UCSI Campus Navigator Application” was successfully developed and tested by the students. The user requirements were successfully embedded into the mobile navigator application and hence met the aims of this research.

Object location determination in a room is a system which can determine the position of an object or person in a physical room. Localization systems which have been built using wireless sensor networks such as RFID, GPS, or Cellular networks are considered less effective. Hence, the present camera usage system as low-cost position determination will be developed. This system consists of two processes which are single-camera processing and multi-camera processing. In single-camera processing, we present object detection using Multi-Scale Spectral Residual (MSR) preprocessing to obtain camera position and best enviroment condition which will be used in multi-camera processing. In multi-camera processing, the object location detection result in the single camera will be combined and tracked in the room floor plan. From the experiment, the result is that human localization efficiency in room is 87.07% in average with minimum error accuracy is 0.02 meter and maximum error accuracy is 2.2 meter.

This paper presents a new approach to recognize elements in floor plan layouts. Besides walls and rooms, we aim to recognize diverse floor plan elements, such as doors, windows and different types of rooms, in the floor layouts. To this end, we model a hierarchy of floor plan elements and design a deep multi-task neural network with two tasks: one to learn to predict room-boundary elements, and the other to predict rooms with types. More importantly, we formulate the room-boundary-guided attention mechanism in our spatial contextual module to carefully take room-boundary features into account to enhance the room-type predictions. Furthermore, we design a cross-and-within-task weighted loss to balance the multi-label tasks and prepare two new datasets for floor plan recognition. Experimental results demonstrate the superiority and effectiveness of our network over the state-of-the-art methods.

The present work describes the algorithmic and hardware engineering perspectives used to build an optimized Automated Guided Vehicle (AGV) robot which can efficiently navigate a warehouse floor plan using a combination of indoor, ultra-wideband wireless positioning sensors, and ultrasonic sensors for obstacle distance measurement and avoidance. The methods behind the trigonometric algorithm modeling are outlined, and several performance tests of the wireless localization system are achieved, both in clear line-of-sight and with adding static, respectively dynamic random interference, in order to verify the high accuracy of the proposed solution.

A rent prediction model with floor plan image (FPI) is proposed in order to evaluate that the image contributes to rent prediction accuracy. The proposed model calculates image feature vectors from FPIs using principal component analysis, combines them and attribute vectors, and applies a regressor. Mean squared errors (MSEs) between predicted and ground truth rents of the proposal are measured using 1,089,090 rental properties of the LIFULL HOME'S dataset. The experimental result suggests that the proposed model with linear regression and FPI slightly improves MSEs compared with the without FPI models; however, the differences are not significant and it is difficult to find merits to use FPIs for rent prediction model.

Graph matching is efficient to retrieve similar layout from the huge size of architectural layout. Since floor plan matching using spectral embedding is popular and its computational time is only in seconds. However, the weakness of this method is the isomorphism of each floor plan that reduces accuracy in the matching process. In this paper, we propose a modified graph representation for room layout matching using spectral embedding. General graph representations of the floor plan assume the node as rooms and the edge as the connection between rooms. In addition, the spectral embedding aims to find the descriptor of each floor plan by ignoring the semantic of rooms. We also consider both room semantic that is the connection between the area outside and inside the room, and the structure of each layout. Moreover, we mathematically proved that adding an extra node can handle the isomorphism of a graph by eigenvalue testing. Our proposed method consists of three processes: i) floor plan extracting, ii) appended topology graph and iii) floor plan matching. The performance from our experiment shows that our proposed method can increase the matched accuracy from the conventional method for about 35.44 percent.

In the recent past, due to the boost in digitization a huge volume of data (BigData) related to building and architectures has been accumulated. The sources of such data are interviews, blogs, websites, etc. These descriptions has details about the interior of a building. For humans it is easy to interpret and imagine its structure and arrangements of furniture. Automatic synthesis of real-world images from text descriptions has been explored in the computer vision community. However, there is no such attempt in the area of document images, like floor plans. Even though, Floor plan synthesis from sketches, as well as data-driven models, were proposed earlier, this is the first attempt to automatically render building floor plan images from textual description. Here, the input is a natural language description of the internal structure and furniture arrangements within a house, and the output is the 2D floor plan image of the same. We have experimented on publicly available benchmark floor plan datasets. We were able to render realistic synthesized floor plan images from the description written in English.

Indoor positioning technologies are subject to serious research in recent years. Starting from low level inertial sensor based, through radio, magnetic or vision based solutions, ending on multiple sensors based fusions. All of the methods mentioned above can be designed in a way that will allow to benefit from vectorized indoor maps. Furthermore in order to be valuable to the end user, output of indoor positioning method have to be presented in a graphical way. Most common is projection of location into indoor floorplan. Nowadays it is common that modern buildings are computer designed and therefore are equipped with precise architectural plans that are necessary to issue building permit and helpful during construction. On the other hand old buildings, that for some reason require renewed floor plans are surveyed and architectural plans are drawn. A wealth of details that are unnecessary from the positioning system point of view, prevent us from using such plans for indoor positioning applications. Indoor maps have to be simplified in order to reduce complexity. This can be achieved by manual or automatic modifications of architectural drawings or raster images of floor plans. This work presents system designed to setup and maintain indoor positioning in a building and novel semi-automatic method for construction of vectorized floor plan from raster images using door detection and flood fill algorithm.

In this paper, a large scale public dataset containing floor plan images and their annotations is presented. BRIDGE (Building plan Repository for Image Description Generation, and Evaluation) dataset contains more than 13000 images of the floor plan and annotations collected from various websites, as well as publicly available floor plan images in the research domain. The images in BRIDGE also has annotations for symbols, region graphs, and paragraph descriptions. The BRIDGE dataset will be useful for symbol spotting, caption and description generation, scene graph synthesis, retrieval and many other tasks involving building plan parsing. In this paper, we also present an extensive experimental study for tasks like furniture localization in a floor plan, caption and description generation, on the proposed dataset showing the utility of BRIDGE.

Localization in an indoor or Global Positioning System (GPS)-denied environment is paramount. It drives various applications that require locating humans or robots in an unknown environment. Various localization systems using different ubiquitous sensors such as camera, radio frequency, inertial measurement unit have been developed. Most of these systems cannot accommodate for scenarios which have substantial changes in the environment such as a large number of people (unpredictable) and sudden change in the environment floor plan (unstructured). In this paper, we propose a system, InFo that can leverage real-time visual information captured by surveillance cameras and augment that with images captured by the smart device user to deliver accurate discretized location information. Through our experiments, we demonstrate that our deep learning based InFo system provides an improvement of 10% as compared to a system that does not utilize this real-time information.

In this paper we present a method for text recognition in floor plan images. In particular, we are concerned about locating, reading, and categorizing text inside floor plan images to obtain information about the building. Furthermore, the aim of this paper is to compare traditional text detection methods, based on image processing techniques, with recent approaches relying on convolutional neural networks. To improve results we combined several methods outperforming the original ones. Text regions are also classified in four semantic classes according to their purpose. Two datasets with different features, including quality and size, were considered in the experiments performed.

Micro Aerial Vehicles (MAVs) lend themselves to perform supporting flights in many cases from consumer-oriented service to Urban Search and Rescue (USAR) scenarios. However, most of today’s approaches are restricted to piloted outdoor applications for lack of autonomy in unknown scenarios. In this paper, we present a comprehensive mapping system tailored to the specific needs of independent indoor reconnaissance flights. We provide for both metric and floor plan mapping and exploit the benefits of online extracted semantics in two ways. First, grid-mapping performance directly benefits from semantic feedback loop. Second, the MAV uses a multi-layered topological map established on the fly for autonomous reconnaissance in unknown environment. A new exploration strategy provides for complete coverage in consideration of desired prioritization. We evaluate the proposed enhancements in robust mapping and floor plan generation for MAVs by means of real experimental data acquired in demanding indoor environment. The flight guidance is validated in simulations.

Accurate indoor positioning is challenging due to non-line-of-sight (NLOS) conditions which cause ranging and location errors. This paper proposes an iterative algorithm to improve location accuracy in NLOS conditions caused by through-the-wall (TTW) propagation. The proposed algorithm uses a floor plan with details of walls’ locations and dimensions. The algorithm reduces the NLOS range between a tag and anchors iteratively, resulting in a final location that is closer to the tag’s true location than the NLOS estimated location. It is shown that the algorithm can reduce the impact of NLOS conditions on location accuracy significantly.

In order to estimate the position of the unmanned aerial vehicle (UAV) in an indoor disaster environment, it is recommended to utilize methods using wireless signals regardless of smoke or dust. Among the wireless signal-based localization methods, UWB-based localization is becoming more popular. In an environment surrounded by walls or pillars, there is a significant difference between UWB signal’s line-of-sight (LOS) measurement and non-line-of-sight (NLOS) measurement. In this paper, we propose the UWB-based localization method that compensates for UWB ranging measurement by predicting NLOS multipath using a ray-tracing algorithm based on given floor plan. Experiments were conducted indoors in the test complex site. The experimental results verifies that the proposed algorithm has improved localization performance compared to the traditional algorithms.

The floor plan is one of the most common ways to display details information about space or residence. However, traditional classifiers such as Logistic Regression (LR) or K-Nearest Neighbor (KNN) cannot effectively read the high-level information from the floor plan. Besides, the training process requires a large number of labeled data which is difficult to collect. However, transfer learning can deal with classification problems by using existing knowledge to solve different but related problems. This paper presents a method to distinguish the structure of floor plans based on convolutional neural network and multi-layer perceptron (MLP). Firstly, the paper uses a famous convolution neural network known as Visual Geometry Group (VGGNet) as a feature extractor to extract features of images. And then design a unique Auto-encoder to reduce the dimensionality of data. Finally, the multi-layer perceptron is intended to achieve the classification job. Compared with traditional machine learning models, this method improves the accuracy of classification and proves the transfer learning actually has a beneficial effect on structured image Classification.

The increasing trend of using online platforms for real estate rent/sale makes automatic retrieval of similar floor plans a key requirement to help architects and buyers alike. Although sketch based image retrieval has been explored in the multimedia community, the problem of hand-drawn floor plan retrieval has been less researched in the past. In this paper, we propose REXplore (Real Estate eXplore), a novel framework that uses sketch based query mode to retrieve corresponding similar floor plan images from a repository using Cyclic Generative Adversarial Networks (Cyclic GAN) for mapping between sketch and image domain. The key contributions of our proposed approach are : (1) a novel sketch based floor plan retrieval framework using an intuitive and convenient sketch query mode; (2) A conjunction of Cyclic GANs and Convolution Neural Networks (CNNs) for the task of hand-drawn floor plan image retrieval. Extensive experimentation and comparison with baseline results authenticates our claim.