Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Lightweight convolutional neural network driven by small data for asphalt pavement crack segmentation
Abstract A lightweight PCSNet-based segmentation model is developed to efficiently overcome insufficient performance in feature extraction and boundary loss information resulting from sampling operations. The proposed approach incorporates two key components: an enhanced shuffle unit and an improved inverted residual architecture to effectively reduce model parameters and enhance inference time. Additionally, introducing the generalized Dice loss (GDL) aims to address the prediction accuracy issue caused by category imbalance. Finally, this study employs gradient-based class activation mapping to visualize and interpret the learned features. To enhance model interpretability, experiments were conducted on three publicly available datasets and compared with existing popular segmentation models. The findings demonstrate that including the GDL in the lightweight PCSNet model leads to a significant improvement in prediction performance, with the highest mIoU reaching 78.93%. Additionally, the visualization of class activation mapping (CAM) further enhances PCSNet interpretability in terms of feature extraction.
Highlights A customizable and lightweight segmentation model improves inference time. The improved inverted residual of the encoder reduces the model's parameters. The enhanced shuffle unit structure improves the model's prediction accuracy. The GDL loss function is introduced to address the category imbalance issue. The interpretability of CNNs is explained from the perspective of feature extraction.
Lightweight convolutional neural network driven by small data for asphalt pavement crack segmentation
Abstract A lightweight PCSNet-based segmentation model is developed to efficiently overcome insufficient performance in feature extraction and boundary loss information resulting from sampling operations. The proposed approach incorporates two key components: an enhanced shuffle unit and an improved inverted residual architecture to effectively reduce model parameters and enhance inference time. Additionally, introducing the generalized Dice loss (GDL) aims to address the prediction accuracy issue caused by category imbalance. Finally, this study employs gradient-based class activation mapping to visualize and interpret the learned features. To enhance model interpretability, experiments were conducted on three publicly available datasets and compared with existing popular segmentation models. The findings demonstrate that including the GDL in the lightweight PCSNet model leads to a significant improvement in prediction performance, with the highest mIoU reaching 78.93%. Additionally, the visualization of class activation mapping (CAM) further enhances PCSNet interpretability in terms of feature extraction.
Highlights A customizable and lightweight segmentation model improves inference time. The improved inverted residual of the encoder reduces the model's parameters. The enhanced shuffle unit structure improves the model's prediction accuracy. The GDL loss function is introduced to address the category imbalance issue. The interpretability of CNNs is explained from the perspective of feature extraction.
Lightweight convolutional neural network driven by small data for asphalt pavement crack segmentation
Liang, Jia (Autor:in) / Zhang, Qipeng (Autor:in) / Gu, Xingyu (Autor:in)
20.11.2023
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Vision-Based Crack Detection of Asphalt Pavement Using Deep Convolutional Neural Network
| Springer Verlag | 2021
Recognition of asphalt pavement crack length using deep convolutional neural networks
| Taylor & Francis Verlag | 2018
Convolutional neural network for pothole detection in asphalt pavement
| Taylor & Francis Verlag | 2021