PavementDamagesG-7

Published: 14 January 2026| Version 2 | DOI: 10.17632/yb65trwbd8.2
Contributors:
,
,
, Eimar Sandoval, Maarten Bassier

Description

This repository contains a dataset (PavementDamagesG-7.zip) focused on the detection and classification of surface damage in flexible pavements. The dataset is designed to support computer vision tasks such as segmentation and classification, especially for research in infrastructure monitoring and automated road condition assessment.

Files

Steps to reproduce

Dataset Metadata Dataset name: PavementDamagesG-7 Location: Cali, Colombia Capture period: 2024–2025 Imaging device: StereoLabs ZED 2i stereo camera Resolution: 1920×1080 pixels Format: RGB images (JPG) with corresponding pixel-wise annotation masks (PNG) Annotation scheme: • Longitudinal/Transverse crack — colorRGB (250, 50, 83) • Alligator cracking — colorRGB (61, 245, 61) • Pothole — colorRGB (42, 125, 209) • Pavement marking — colorRGB (245, 147, 49) • Drains — colorRGB (115, 51, 128) • Patching — colorRGB (250, 250, 55) • Background / No damage — colorRGB (0, 0, 0) Description: The images were captured across multiple road segments in Cali, Colombia, under varying lighting and environmental conditions to ensure visual diversity. The ZED 2i camera enabled high-resolution stereo imaging, facilitating annotation. The dataset was curated between 2024 and 2025 as part of a research initiative on automated pavement surface evaluation. Citation: If you use this dataset, please cite: Tello-Cifuentes, L., Thomson, P., Marulanda, J., Sandoval, E., & Bassier, M. (2026). Deep learning-based detection and evaluation of pavement surface damage. Results in Engineering, 29, 108946. https://doi.org/10.1016/j.rineng.2025.108946

Institutions

  • Universidad del Valle
  • Katholieke Universiteit Leuven

Categories

Image Segmentation, Object Detection, Surface Damage, Convolutional Neural Network, Damage Classification

Funders

  • Ministerio de Ciencia, Tecnología e Innovación- Colombia
    Grant ID: code CI 100613, 933–2023

Licence