Related Experiment Video
Updated: May 2, 2026
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
996
LiteCrackSeg: A lightweight hybrid CNN-transformer for efficient crack segmentation
Kaleb Amsalu Gobena1, Md Youshuf Khan Rakib1, Fiseha Berhanu Tesema2
1School of Computer Science and Engineering, Central South University, Changsha, Hunan, China.
Plos One
|April 30, 2026
Summary
LiteCrackSeg, a new lightweight AI model, accurately detects infrastructure cracks using a hybrid CNN-transformer approach. This efficient system enables real-time structural health monitoring on edge devices.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Structural Engineering
Background:
- Infrastructure cracks are key indicators of structural deterioration.
- Automated crack segmentation is vital for structural health monitoring.
- Challenges include thin, low-contrast cracks and class imbalance.
Purpose of the Study:
- To develop an efficient and accurate crack segmentation model for resource-constrained devices.
- To address the challenges of crack morphology and class imbalance in segmentation.
Main Methods:
- Proposed LiteCrackSeg, a lightweight hybrid CNN-transformer architecture.
- Utilized a MobileViT encoder for local and global feature extraction.
- Introduced a Morphology-Aware MobileViT (MAM-ViT) bottleneck with Dynamic Snake Convolutions (DSConv).
- Employed a transformer-based decoder and attention-guided fusion.
- Trained using Tversky loss to handle class imbalance.
Main Results:
- Achieved state-of-the-art segmentation performance on DeepCrack, CrackMap, and TUT datasets.
- Demonstrated high computational efficiency with 2.72M parameters and 3.23 GFLOPs.
- Real-time inference at 56 FPS on 512x512 images.
Conclusions:
- LiteCrackSeg offers an efficient and accurate solution for infrastructure crack segmentation.
- The model's lightweight design is suitable for deployment on edge devices for practical inspection.
- Enables advanced structural health monitoring in real-world applications.
Related Concept Videos
01:20Microcracking in Concrete
614
Microcracking in concrete refers to the tiny cracks that can form within the material even before any external load is applied. These microcracks typically occur at the interface between the coarse aggregate and the hydrated cement paste, often as a result of differential volume changes prompted by variations in stress-strain behavior, as well as thermal and moisture movement. Initially, these microcracks remain stable and do not grow substantially until the concrete is stressed to about 30...
614
01:28Types of Non-structural Cracks in Concrete
668
Non-structural cracks are primarily of three types: plastic, early-age thermal, and drying shrinkage cracks. Plastic cracks are further classified into plastic shrinkage cracks and plastic settlement cracks.
Plastic shrinkage cracks typically form within hours after the concrete is poured. The concrete's surface dries faster than the bottom, creating tensile stress that the still-plastic concrete cannot withstand, leading to diagonal or randomly patterned cracks on the concrete surface.
Plastic shrinkage cracks typically form within hours after the concrete is poured. The concrete's surface dries faster than the bottom, creating tensile stress that the still-plastic concrete cannot withstand, leading to diagonal or randomly patterned cracks on the concrete surface.
668