A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Segmentation of Concrete Surface Cracks Using DeeplabV3+ Architecture
Concrete is a common construction material used in structural engineering, but it is prone to cracks which can negatively impact the quality and longevity of structures. Therefore, timely and accurate detection of cracks in concrete surfaces is an important task in structural health monitoring. Currently, deep learning has emerged as a powerful technique in different fields due to its ability to learn from large data sets, recognize patterns, and make accurate predictions. The aim of this study is to suggest an effective backbone solution for the concrete surface crack detection task using DeepLabv3+ architecture. Specifically, seven different back-bones investigated in this study were MobileNet-v2, EfficientNet-b0, Res-NeXt50-32x4d, timm-regNetx-002, timm-regNety-002, timm-gerNet-s, timm-efficientNet-b0. For the training process, we used the Adam algorithm for updating the weights of the model and the Dice loss function as the objective function. The study results show that all backbones effectively detected concrete cracks with over 92% Intersection over Union (IoU). The ResNeXt50-32x4d presents the best performance of 93.8% IoU. The findings highlighted the feasibility and effectiveness of models in concrete crack segmentation tasks.
Segmentation of Concrete Surface Cracks Using DeeplabV3+ Architecture
Concrete is a common construction material used in structural engineering, but it is prone to cracks which can negatively impact the quality and longevity of structures. Therefore, timely and accurate detection of cracks in concrete surfaces is an important task in structural health monitoring. Currently, deep learning has emerged as a powerful technique in different fields due to its ability to learn from large data sets, recognize patterns, and make accurate predictions. The aim of this study is to suggest an effective backbone solution for the concrete surface crack detection task using DeepLabv3+ architecture. Specifically, seven different back-bones investigated in this study were MobileNet-v2, EfficientNet-b0, Res-NeXt50-32x4d, timm-regNetx-002, timm-regNety-002, timm-gerNet-s, timm-efficientNet-b0. For the training process, we used the Adam algorithm for updating the weights of the model and the Dice loss function as the objective function. The study results show that all backbones effectively detected concrete cracks with over 92% Intersection over Union (IoU). The ResNeXt50-32x4d presents the best performance of 93.8% IoU. The findings highlighted the feasibility and effectiveness of models in concrete crack segmentation tasks.
Segmentation of Concrete Surface Cracks Using DeeplabV3+ Architecture
Lecture Notes in Civil Engineering
Reddy, J. N. (editor) / Wang, Chien Ming (editor) / Luong, Van Hai (editor) / Le, Anh Tuan (editor) / Nguyen, Tan-No (author) / Tran, Thanh Danh (author) / Cuong, Phan Viet (author)
The International Conference on Sustainable Civil Engineering and Architecture ; 2023 ; Da Nang City, Vietnam
Proceedings of the Third International Conference on Sustainable Civil Engineering and Architecture ; Chapter: 164 ; 1531-1538
2023-12-12
8 pages
Article/Chapter (Book)
Electronic Resource
English
Semantic Segmentation of Cracks Using DeepLabv3+
| Springer Verlag | 2023
High-precision segmentation and quantification of tunnel lining crack using an improved DeepLabV3+
| Elsevier | 2025
Image Segmentation Method for Sweetgum Leaf Spots Based on an Improved DeeplabV3+ Network
| DOAJ | 2022
Sweetgum Leaf Spot Image Segmentation and Grading Detection Based on an Improved DeeplabV3+ Network
| DOAJ | 2023