This article presents BDRoad-Sense, a systematically collected and validated multi-class road surface anomaly dataset developed to support road safety enhancement and automated road condition classification. The dataset comprises five categories: Major Damage, Minor Damage, Normal Road, Manhole, and Speed Breaker, representing both structural road deterioration and functional infrastructure components. Image collection was conducted from November 7, 2025, to February 16, 2026, across rural and urban regions of Sylhet District, Bangladesh, using four different smartphone cameras to incorporate natural variations in imaging conditions and device characteristics. Initially, 6,350 road surface images were collected and manually screened to ensure class relevance and visual clarity. To improve dataset balance and enhance intra-class variability, controlled data augmentation techniques were applied, generating 2757 augmented images and resulting in a total dataset size of 9,107 images. All images were resized to a uniform resolution of 1024 × 1024 pixels to maintain consistency and compatibility with machine learning and computer vision frameworks. The dataset supports benchmarking of multi-class classification models, including convolutional and transformer-based architectures, and provides a structured resource for automated road monitoring, infrastructure assessment, and intelligent transportation research.
Mashrafi et al. (Fri,) studied this question.