Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

An Improved ResNet50 Model for Predicting Pavement Condition Index (PCI) Directly from Pavement Images

Andrews Danyo, Anthony Dontoh, Armstrong Aboah

TL;DR

This work tackles automatic Pavement Condition Index (PCI) estimation directly from pavement images using an enhanced ResNet50 integrated with a Convolutional Block Attention Module (CBAM). By embedding channel and spatial attention, the model focuses on distress-related features without requiring manual PCI annotations, achieving improved predictive accuracy over baseline ResNet50 and DenseNet161 on the DSPS dataset. The results demonstrate lower error metrics and interpretable attention maps that illuminate the model's reasoning, highlighting the practical potential for scalable automated pavement assessment. The approach advances automated pavement management by reducing labeling costs and enabling real-time condition monitoring across diverse road conditions.

Abstract

Accurately predicting the Pavement Condition Index (PCI), a measure of roadway conditions, from pavement images is crucial for infrastructure maintenance. This study proposes an enhanced version of the Residual Network (ResNet50) architecture, integrated with a Convolutional Block Attention Module (CBAM), to predict PCI directly from pavement images without additional annotations. By incorporating CBAM, the model autonomously prioritizes critical features within the images, improving prediction accuracy. Compared to the original baseline ResNet50 and DenseNet161 architectures, the enhanced ResNet50-CBAM model achieved a significantly lower mean absolute percentage error (MAPE) of 58.16%, compared to the baseline models that achieved 70.76% and 65.48% respectively. These results highlight the potential of using attention mechanisms to refine feature extraction, ultimately enabling more accurate and efficient assessments of pavement conditions. This study emphasizes the importance of targeted feature refinement in advancing automated pavement analysis through attention mechanisms.

An Improved ResNet50 Model for Predicting Pavement Condition Index (PCI) Directly from Pavement Images

TL;DR

This work tackles automatic Pavement Condition Index (PCI) estimation directly from pavement images using an enhanced ResNet50 integrated with a Convolutional Block Attention Module (CBAM). By embedding channel and spatial attention, the model focuses on distress-related features without requiring manual PCI annotations, achieving improved predictive accuracy over baseline ResNet50 and DenseNet161 on the DSPS dataset. The results demonstrate lower error metrics and interpretable attention maps that illuminate the model's reasoning, highlighting the practical potential for scalable automated pavement assessment. The approach advances automated pavement management by reducing labeling costs and enabling real-time condition monitoring across diverse road conditions.

Abstract

Accurately predicting the Pavement Condition Index (PCI), a measure of roadway conditions, from pavement images is crucial for infrastructure maintenance. This study proposes an enhanced version of the Residual Network (ResNet50) architecture, integrated with a Convolutional Block Attention Module (CBAM), to predict PCI directly from pavement images without additional annotations. By incorporating CBAM, the model autonomously prioritizes critical features within the images, improving prediction accuracy. Compared to the original baseline ResNet50 and DenseNet161 architectures, the enhanced ResNet50-CBAM model achieved a significantly lower mean absolute percentage error (MAPE) of 58.16%, compared to the baseline models that achieved 70.76% and 65.48% respectively. These results highlight the potential of using attention mechanisms to refine feature extraction, ultimately enabling more accurate and efficient assessments of pavement conditions. This study emphasizes the importance of targeted feature refinement in advancing automated pavement analysis through attention mechanisms.

Paper Structure

This paper contains 24 sections, 6 equations, 10 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Literature Works
  3. Evolution of PCI Assessment Methods
  4. Shallow Machine Learning Approaches
  5. Advances in Deep Learning for PCI Estimation
  6. Addressing Limitations with Attention Mechanisms
  7. The Gap in Literature and Proposed Solution
  8. Methodology
  9. ResNet50
  10. DenseNet161
  11. Convolutional Block Attention Module
  12. Channel Attention Module
  13. Spatial Attention Module
  14. Proposed ResNet50 + CBAM (Ours)
  15. Experiments
  16. ...and 9 more sections

Figures (10)

  • Figure 1: A diagram of ResNet50 architecture
  • Figure 2: Archtecture for DenseNet161 model
  • Figure 3: Overview of the Convolutional Block Attention Module (CBAM)
  • Figure 4: Proposed architecture for ResNet50-CBAM model
  • Figure 5: Feature map visualization of results
  • ...and 5 more figures