Multiclass Road Surface Damage Classification Dataset: Computer Vision

This dataset contains 1,530 high-resolution, three-channel (RGB) images of road surfaces. It is designed to support machine learning and computer vision research for robust, three-class road damage classification, particularly contributing to innovations in automated infrastructure inspection, predictive maintenance, and smart city frameworks. Dataset Composition: The dataset is organized into three distinct, structurally meaningful categories based on pavement degradation. ⦁ Surface Erosion (651 images): Images displaying weathering, raveling, or degradation of the top asphalt layer. ⦁ Crack (478 images): Images exhibiting linear, transverse, or interconnected (alligator) cracks on the road surface. ⦁ Pothole (401 images): Images showing bowl-shaped depressions or significant missing pavement chunks. Preprocessing: ⦁ Images are harmonized to 640 x 640 pixels and loaded as three-channel RGB images. ⦁ A unified preprocessing pipeline was applied, including resizing, aspect-ratio preservation (via optional zero-padding), and per-image color normalization to ensure consistency across varying lighting conditions. ⦁ The dataset is pre-structured to support standard machine learning workflows, typically divided into an 80/10/10 (train/validation/test) split to ensure robust model evaluation and prevent data leakage. This dataset can be effectively used for: ⦁ Multiclass image classification and automated road anomaly recognition tasks. ⦁ Deep learning model development, including Vision Transformers (ViTs), parameter-efficient CNNs, and real-time edge computing architectures. ⦁ Benchmarking automated pavement distress detection systems for municipal asset management. ⦁ Evaluating autonomous vehicle perception systems for obstacle avoidance and path planning. File Information: ⦁ Total Images: 1,530 ⦁ Image Format: RGB (3-channel) images ⦁ Resolution: 640 x 640 pixels ⦁ Data Structure: Organized into distinct directories based on damage class.