Conference Paper

Improving 2D Construction Plans with Cycle-Consistent Generative Adversarial Networks

Authors:
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

Digital methods such as Building Information Modeling (BIM) offer significant potential in the operation phase of a building but, therefore, require digital as-built models of the existing structures as a prerequisite. The existing building records have to be transferred into a digital model. For this retro-digitization purpose, Computer Vision (CV) and Machine learning (ML) are essential technologies. Thereby, 2D construction plans are an indispensable data basis for these methods to extract the geometry of the existing structures. The extraction requires a highly reliable detection of lines and texts. However, the performance of the CV and ML methods is highly dependent on the quality of the data. Anomalies like discolorations, stains, and fold lines can negatively affect the detection. This is especially case for old, hand-drawn paper plans, which are often the only data source available for old buildings. To integrate old plans better in the detection process by CV and ML, a quality recovery of the plans is necessary. To achieve this, we propose the use of the cycle-consistent Generative Adversarial Network (CycleGAN) that enables style transformation with unpaired data. Hereby transformation means the removal of the stated anomalies. Our results of CycleGAN improved 2D plans show that both text and edge detection methods perform better.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
Among various building information model (BIM) reconstruction methods for existing building, image-based method can identify building components from scanned as-built drawings and has won great attention due to its lower cost, less professional operators and better reconstruction performance. However, this kind of method will cost a great deal of time to design and extract features. Moreover, the manually extracted features have poor robustness and contain less non-geometric information. In order to solve this problem, this paper proposes a deep learning-based method to detect building components from scanned 2D drawings. Taking structural drawings as an example, in this article, 1500 images of structural drawings were firstly collected and preprocessed to guarantee the quality of data. After that, the neural network model—You Only Look Once (YOLO) was trained, verified and tested. In addition, a series of metrics were utilized to evaluate the performance of recognition. The results of test experiments show that the components in structural drawings (e.g., grid reference, column and beam) can be successfully detected, while the average detection accuracy of the whole image is over 80% and the average detection time for each image is 0.71 s. The experimental results demonstrate that the proposed method is robust and timesaving, which provides a good basis for the reconstruction of BIM from 2D drawings.
Article
Full-text available
Building Information Models (BIMs) are valuable facility management tools, however the resources required to develop existing building BIMs are a recognized barrier to adoption of facility management enabled BIMs. Significant developments have been made for generating geometrically complex models using scanning technologies, however the resultant models are often extremely large and require significant computational resources. This paper presents an automated process that uses 2D floorplans and elevation drawings to generate semantically-rich BIMs with adequate geometry for energy simulation and integration of the significant diversity and volume of data necessary for day-to-day building operations and facility management. The proposed approach minimizes the resources necessary for both model creation and maintenance, limiting initial BIM creation to match the level of detail available on available building drawings, while providing the flexibility to incorporate complex geometry at a future date. Two case studies, a set of university campus buildings and a healthcare training facility, have been modelled using this approach and the resultant model generation speed, accuracy, and resultant quality are discussed, along with limitations and future process improvements.
Article
The advances in mobile technology and portable cameras have facilitated enormously the acquisition of text images. However, the blur caused by camera shake or out-of-focus problems may affect the quality of acquired images and their use as input for optical character recognition (OCR) or other types of document processing. This work proposes an end-to-end model for document deblurring using cycle-consistent adversarial networks. The main novelty of this work is to achieve blind document deblurring, i.e., deblurring without knowledge of the blur kernel. Our method, named “Blur2Sharp CycleGAN,” generates a sharp image from a blurry one and shows how cycle-consistent generative adversarial networks (CycleGAN) can be used in document deblurring. Using only a blurred image as input, we try to generate the sharp image. Thus, no information about the blur kernel is required. In the evaluation part, we use peak signal to noise ratio (PSNR) and structural similarity index (SSIM) to compare the deblurring images. The experiments demonstrate a clear improvement in visual quality with respect to the state-of-the-art using a dataset of text images.
Article
A large proportion of indoor spatial data is generated by parsing floor plans. However, a mature and automatic solution for generating high-quality building elements (e.g., walls and doors) and space partitions (e.g., rooms) is still lacking. In this study, we present a two-stage approach to indoor mapping and modeling (IMM) from floor plan images. The first stage vectorizes the building elements on the floor plan images and the second stage repairs the topological inconsistencies between the building elements, separates indoor spaces, and generates indoor maps and models. To reduce the shape complexity of indoor boundary elements, i.e., walls and openings, we harness the regularity of the boundary elements and extract them as rectangles in the first stage. Furthermore, to resolve the overlaps and gaps of the vectorized results, we propose an optimization model that adjusts the rectangle vertex coordinates to conform to the topological constraints. Experiments demonstrate that our approach achieves a considerable improvement in room detection without conforming to Manhattan World Assumption. Our approach also outputs instance-separate walls with consistent topology, which enables direct modeling into Industry Foundation Classes (IFC) or City Geography Markup Language (CityGML).
Article
Recent advances in deep convolutional neural networks (DCNNs) and generative adversarial networks (GANs) have significantly improved the performance of single image blind deblurring algorithms. However, most of the existing algorithms require paired training data. In this paper, we present an unsupervised method for single-image deblurring without paired training images. We introduce a disentangled framework to split the content and blur features of a blurred image, which yields improved deblurring performance. To handle the unpaired training data, a blurring branch and the cycle-consistency loss are added to guarantee that the content structures of the restored results match the original images. We also add a perceptual loss to further mitigate the artifacts. For natural image deblurring, we introduce a color loss to reduce color distortions in outputs. Extensive experiments on both domain-specific and natural image deblurring show the proposed method achieves competitive results compared to recent state-of-the-art deblurring approaches.
Conference Paper
There is large consent that successful training of deep networks requires many thousand annotated training samples. In this paper, we present a network and training strategy that relies on the strong use of data augmentation to use the available annotated samples more efficiently. The architecture consists of a contracting path to capture context and a symmetric expanding path that enables precise localization. We show that such a network can be trained end-to-end from very few images and outperforms the prior best method (a sliding-window convolutional network) on the ISBI challenge for segmentation of neuronal structures in electron microscopic stacks. Using the same network trained on transmitted light microscopy images (phase contrast and DIC) we won the ISBI cell tracking challenge 2015 in these categories by a large margin. Moreover, the network is fast. Segmentation of a 512x512 image takes less than a second on a recent GPU. The full implementation (based on Caffe) and the trained networks are available at http://lmb.informatik.uni-freiburg.de/people/ronneber/u-net .
Article
The overall energy efficiency of existing buildings has to be significantly improved to comply with emerging regulations and to contribute to overcoming current environmental challenges. Many policies aim at accelerating the renovation rate. The effectiveness of renovation actions could be significantly improved through the systematic use of Information and Communication Technologies (ICT) tools and Building Information Modeling (BIM). But these solutions rely on full-fledged digital models, which, for most buildings, are not available. The present article introduces a research work aiming at the development of methods for the generation of 3D building models from 2D plans. The developed prototype is able to extract information from 2D plans and to generate IFC (Industry Foundation Classes)-compliant 3D models that include the main components of the building: walls, openings, and spaces. The article also presents the results of a quantitative assessment of the platform capabilities and performances, relying on a database of 90 real architectural floor plans. The results are very promising and show that such solutions could be key components of future digital toolkits for renovation design.
Indoor mapping and modeling by parsing floorplan images
  • Y Wu
  • J Shang
  • P Chen
  • X Hu
  • Z Zhou
  • Wu Y.
  • I J Goodfellow
  • J Pouget-Abadie
  • M Mirza
  • B Xu
  • D Warde-Farley
  • S Ozair
  • A Courville
  • Y Bengio
Goodfellow, I. J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A. and Bengio, Y. (2014). "Generative adversarial networks". arXiv preprint arXiv:1406.2661.
  • X Luo
  • X Zhang
  • P Yoo
  • R Martin-Brualla
  • J Lawrence
  • S Seitz
Luo, X., Zhang, X., Yoo, P., Martin-Brualla, R., Lawrence, J., & Seitz, S. (2020). "Time-Travel Rephotography". ArXiv, abs/2012.12261