BDRS2025: Road Surface Image and Video Datasets from Bangladesh.

BDRS2025, a labeled image and video dataset collected from real road environments in Bangladesh for the purpose of road condition analysis using deep learning methods. The dataset contains 2,685 smartphone-captured images and 147 supplementary video data representing five categories of road conditions and road-related signs, such as pothole, crack, speed breaker, zebra crossing, and manhole cover. Images and videos were collected from multiple locations, including Dhaka, Chandpur, and Lakshmipur, under different weather, lighting, and traffic conditions. Each class contains an average of 500 annotated images. The images were manually cleaned and labeled to ensure consistency and accuracy. The dataset provides high-resolution road surface imagery captured in natural driving environments, making it suitable for computer vision research, infrastructure monitoring systems, and intelligent transportation applications. This dataset can be reused for tasks such as road damage classification, object detection, semantic segmentation, and benchmarking deep learning models designed for road condition monitoring in developing countries.