Deep Floor Plan Recognition Using a Multi-Task Network With Room-Boundary-Guided Attention

Zeng, Zhiliang; Li, Xianzhi; Yu, Ying; Fu, Chi‐Wing

doi:10.1109/iccv.2019.00919

Cited by 87 publications

(109 citation statements)

References 21 publications

Supporting

Mentioning

108

Contrasting

Order By: Relevance

“…The proposed framework outputs the result in a vector format, which facilitates its use in additional research or real-world applications. For example, Zeng et al [10] demonstrated the 3D models of the results from their method, and the output of the proposed framework is already vector-type data, making it even easier for 3D modeling.…”

Section: Discussionmentioning

confidence: 99%

“…They also used OCR to be able to recognize the size of the rooms and to place furniture models scaled to the scene. Zeng et al [10] proposed a method that detects and classifies walls, doors, windows, and rooms by training a VGG encoder-decoder. Unlike [8], their method is applicable to non-rectangular shape elements and is able to obtain the shape features of indoor elements.…”

Section: Rule-based Heuristic Methods and Machine Learning Algorithmsmentioning

confidence: 99%

“…In many cases, room detection and recognition depend heavily on structural elements in the floor plan, such as walls, doors, and/or windows, and if these structural elements have unclosed issues, they will considerably affect the room formation process. Because elements may lose their shape information during the vectorization process [7,8], some approaches omit this process in order to secure the shape features [9,10]. In addition, none of the approaches that concentrated on detecting structural elements and space elements considers symbolic elements, such as cabinets, baths, or toilets, among others.…”

Section: Rule-based Heuristic Methods and Machine Learning Algorithmsmentioning

confidence: 99%

“…CNN-based approaches only require a basic level of image pre-processing techniques and are robust to floor plan noise. In addition, they can be applied to any style of drawing without the need for transformation, which makes them efficient and versatile [7][8][9][10].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Framework for Indoor Elements Classification via Inductive Learning on Floor Plan Graphs

Song

2021

IJGI

View full text Add to dashboard Cite

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.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Rule-based Heuristic Methods and Machine Learning Algorithmsmentioning

confidence: 99%

Section: Rule-based Heuristic Methods and Machine Learning Algorithmsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Framework for Indoor Elements Classification via Inductive Learning on Floor Plan Graphs

Song

2021

IJGI

View full text Add to dashboard Cite

show abstract

“…However, this algorithm lacks robustness for unseen elements. A deep multi-task neural network was proposed to identify room types and walls (Zeng et al 2019). Compared with the Raster-to-Vector algorithm (Liu et al 2017), this algorithm has higher precision.…”

Section: Ai-based Bim Model Generationmentioning

confidence: 99%

Building a Next Generation AI Platform for AEC: A Review and Research Challenges

Wang¹,

He²,

Yu³

et al. 2020

CIB W78 Proceedings

View full text Add to dashboard Cite

Transformation of the AEC industry has been found to face significant and persistent hurdles both internal and external to the industry, such as industrial fragmentation, difficulty in talent recruiting, inadequate collaboration and knowledge transfer. Meanwhile, advances and maturity in digital technologies recent years offer tremendous potential to tackle these challenges. Examples include Artificial Intelligence (AI) and deep learning, computer vision, big data analytics, Building Information Models (BIM), 3D scanning, Augmented Reality (AR), Unmanned Aerial Vehicles (UAVs), autonomous driving, and blockchain. We envision that the next generation platform for the AEC industry shall integrate this diverse list of digital technologies. In this paper, we review the stateof-the-art in related literature aiming at creating such an integrated platform. Our review is organized into the following four themes, and for each of them, we also identify gaps and challenges. 1. AI-based BIM model generation: this topic addresses how to make BIM models more ubiquitous, which includes the automatic generation of As-planned BIM model from 2D drawings and the automatic generation of As-built BIM model from site images/videos and point clouds. 2. Blockchain-based collaboration: this theme explores the possibilities of applying Blockchain for AEC to improve various essential aspects such as trust in multi-party collaboration and contract enforcement. 3. AR-based visualization: this topic studies the AR-based visualization for construction management, which aims to increase the efficiency of information transmission. 4. AI-based Construction management: this area reviews the application of AI in progress monitoring, safety management, quality management, and contract management. Based on the above analysis, we conclude by outlining potential topics of a research framework for creating a next generation AI platform for AEC.

show abstract

Automating computational design with generative AI

Ploennigs,

Berger

2024

Civil Engineering Design

View full text Add to dashboard Cite

AI image generators based on diffusion models have recently garnered attention for their capability to create images from simple text prompts. However, for practical use in civil engineering they need to be able to create specific construction plans for given constraints. This paper investigates the potential of current AI generators in addressing such challenges, specifically for the creation of simple floor plans. We explain how the underlying diffusion‐models work and propose novel refinement approaches to improve semantic encoding and generation quality. In several experiments we show that we can improve validity of generated floor plans from 6% to 90%. Based on these results we derive future research challenges considering building information modeling. With this we provide: (i) evaluation of current generative AIs; (ii) propose improved refinement approaches; (iii) evaluate them on various examples; (iv) derive future directions for diffusion models in civil engineering.

show abstract

Deep Floor Plan Recognition Using a Multi-Task Network With Room-Boundary-Guided Attention

Cited by 87 publications

References 21 publications

Framework for Indoor Elements Classification via Inductive Learning on Floor Plan Graphs

Framework for Indoor Elements Classification via Inductive Learning on Floor Plan Graphs

Building a Next Generation AI Platform for AEC: A Review and Research Challenges

Automating computational design with generative AI

Contact Info

Product

Resources

About