PavAnalytics

Automating Pavement Condition Rating: A Deep Learning-Based Framework for Cycle Routes and Greenways.

Overview

PavAnalytics is a deep learning framework designed to automatically assess pavement surface conditions from cycling infrastructure imagery.

The project includes:

• Fisheye Distortion Correction
• Pavement Condition Classification
• Explainable AI (Grad-CAM)
• Performance Evaluation

Project Website

For additional project information, demonstrations, and results, visit:

Project Website: [https://www.paveanalytics.eu/]

Publications

Conference Papers

• Shah, S. M. H., Qureshi, W. S., Dea, G. O. and Ullah, I. (2025). Intelligent Pavement Condition Rating System for Cycle Routes and Greenways. In Proceedings of the 11th International Conference on Vehicle Technology and Intelligent Transport Systems - VEHITS; ISBN 978-989-758-745-0; ISSN 2184-495X, SciTePress, pages 668-675. DOI: 10.5220/0013505000003941
• Garcia, J. A. A., Shah, S. M. H., Baig, M. H., Qureshi, W. S. and Ullah, I. (2025). Enhancing Pavement Condition Assessment: A Comprehensive Review of Affordable Sensing Technologies for Cycle Tracks. In Proceedings of the 11th International Conference on Vehicle Technology and Intelligent Transport Systems - VEHITS; ISBN 978-989-758-745-0; ISSN 2184-495X, SciTePress, pages 676-682. DOI: 10.5220/0013505100003941
• M. H. Baig, J. A. Ayala Garcia, W. S. Qureshi, I. Ullah, Towards assessing cycleway pavement surface roughness using an action camera with imu and gps, in Proceedings of the 11th International Conference on Vehicle Technology and Intelligent Transport Systems - VEHITS, INSTICC, SciTePress, 2025, pp. 247–255. doi:10.5220/0013504900003941.

Pavement Condition Rating Scale

We have introduced an intelligent pavement rating system for cycleways, named the Cycle Route Surface Index, a colour-coded, five-level rating system that combines visual inspection, roughness, vegetation, and drainage data to provide a clear and consistent pavement quality measure.

Project Structure

PavAnalytics/
├── data/                  # Dataset and sample images
├── fisheye_correction/    # Fisheye distortion correction pipeline
├── classification/        # Pavement condition classification models
├── explainability/        # Grad-CAM and XAI visualisations
├── figures/               # rating-scale Rubric image
└── README.md

Technology Stack

Core Libraries

Library / Framework	Purpose
Python	Core programming language
PyTorch	Deep learning model development and training
Hugging Face	Access to pretrained models and transformer tools
Transformers	Swin Transformer implementation
OpenCV	Image processing and fisheye distortion correction
NumPy	Numerical computations
Pandas	Data handling and analysis
Matplotlib	Visualisation and result analysis
Scikit-learn	Performance evaluation metrics
Pillow (PIL)	Image loading and processing
Grad-CAM	Explainable AI visualisations

Status

The repository is currently under development. Source code, model checkpoints, installation instructions, and reproducibility documentation will be released following publication of the associated research papers.