A Deep Learning-Based Method to Detect Components from Scanned Structural Drawings for Reconstructing 3D Models

Zhao, Yunfan; Deng, Xueyuan; Lai, Huahui

doi:10.3390/app10062066

Cited by 27 publications

(18 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Faster R-CNN includes an end-to-end model, and region proposal-based CNN architectures for object detection are often used to identify the parts and components in floor plans [1,3]. The YOLO model, which is a fast one-stage convolutional method that automatically detects structural components from scanned CAD drawings, was proposed in [21]. Several improved versions of the YOLO model, such as YOLOv2, have also been applied extensively [22].…”

Section: Recognition Tasksmentioning

confidence: 99%

MuraNet: Multi-task Floor Plan Recognition with Relation Attention

Huang¹,

Wu²,

Wei³

et al. 2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

The recognition of information in floor plan data requires the use of detection and segmentation models. However, relying on several single-task models can result in ineffective utilization of relevant information when there are multiple tasks present simultaneously. To address this challenge, we introduce MuraNet, an attention-based multitask model for segmentation and detection tasks in floor plan data. In MuraNet, we adopt a unified encoder called MURA as the backbone with two separated branches: an enhanced segmentation decoder branch and a decoupled detection head branch based on YOLOX, for segmentation and detection tasks respectively. The architecture of MuraNet is designed to leverage the fact that walls, doors, and windows usually constitute the primary structure of a floor plan's architecture. By jointly training the model on both detection and segmentation tasks, we believe MuraNet can effectively extract and utilize relevant features for both tasks. Our experiments on the CubiCasa5k public dataset show that MuraNet improves convergence speed during training compared to single-task models like U-Net and YOLOv3. Moreover, we observe improvements in the average AP and IoU in detection and segmentation tasks, respectively. Our ablation experiments demonstrate that the attention-based unified backbone of MuraNet achieves better feature extraction in floor plan recognition tasks, and the use of decoupled multi-head branches for different tasks further improves model performance. We believe that our proposed Mu-raNet model can address the disadvantages of single-task models and improve the accuracy and efficiency of floor plan data recognition.

show abstract

Section: Recognition Tasksmentioning

confidence: 99%

MuraNet: Multi-task Floor Plan Recognition with Relation Attention

Huang¹,

Wu²,

Wei³

et al. 2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…All detections are merged to enable the utilization of Dynamo to create the digital model. Similarly, Zhao et al (2020) uses the YOLO object detector to locate structural elements in the column structure and generate framework plan images. In a subsequent study, Zhao et al (2021) continues the research by incorporating the superior Faster R-CNN model and introducing the creation of an IFC file from the extracted information.…”

Section: Related Workmentioning

confidence: 99%

Automated Extraction of Bridge Gradient from Drawings Using Deep Learning

Bayer,

Faltin,

König

2023

CONVR 2023 - Proceedings of the 23rd International Conference on Construction Applications of Virtual Reality

View full text Add to dashboard Cite

Digital methods such as Building Information Modeling (BIM) can be leveraged, to improve the efficiency of maintenance planning of bridges. However, this requires digital building models, which are rarely available. Consequently, these models must be created retrospectively, which is time-consuming when done manually. Naturally, there is a great interest in the industry to automate the process of retro-digitization. This paper contributes to these efforts by proposing a multistage pipeline to automatically extract the gradient of a bridge from pixel-based construction drawings using deep learning. The bridge gradient, a key element of the structure’s axis, is critical for describing the elevation profile and axis slope. This information is implicitly contained in the longitudinal view of bridge drawings as gradient symbols. To extract this information, the well-established object detection model YOLOv5 is employed to locate the gradient symbols inside the drawings. Subsequently, EasyOCR and heuristic rules are applied to extract the relevant gradient parameters associated with each detected symbol. The extracted parameters are then exported in a machine-interpretable format to facilitate seamless integration into other applications. The results show a promising 98% accuracy in symbol detection and an overall accuracy of 70%. Consequently, the pipeline represents a significant advance in automating the retro-digitization process for existing bridges by reducing the time and effort required

show abstract

“…The second paper, by M. Hamooni, M. Maghrebi, J. M. Sardroud and S. Kim presents a novel method of monitoring the maturity of concrete and providing reduced formwork removal time with the strength ensured in real-time [13]. The last one, authored by Y. Zhao, X. Deng, and H. Lai proposes a deep learning-based method [14].…”

Section: Bim In the Construction Industrymentioning

confidence: 99%