BD-PaveSurface: A Multi-Weather Pavement Surface Image Dataset for Crack, Pothole, and Good Pavement Condition Assessment in Bangladesh

BD-PaveSurface is a field-collected pavement surface image dataset developed for image-level assessment of three pavement surface conditions: Crack, Pothole, and Good pavement surface. The images were collected from selected sections of National Highways N5 and N6 in Pabna District, Bangladesh, using smartphone cameras under natural outdoor and real traffic-environment conditions across rainy-season, winter, and summer field surveys. The dataset contains 21,000 stored image files, including 9,000 cleaned original field images (3,000 per class) and 12,000 processed/augmented files comprising standardized 224 × 224 RGB versions of the 9,000 cleaned original images and 3,000 newly generated training-only augmented images (1,000 per class). The cleaned original images were assigned to training, validation, and testing subsets using a similarity-controlled scene-level 70:15:15 split; after training-only augmentation, the processed dataset contains 9,300 training images, 1,350 validation images, and 1,350 testing images, with no augmentation applied to the validation or testing subsets. The repository also provides image-level metadata, scene-level split assignments, SHA-256 cross-split duplicate-audit results, a reproducibility notebook, software-environment information, and reuse-license documentation to support pavement-condition classification, crack and pothole recognition, transfer learning, augmentation studies, computer-vision benchmarking, and camera-based road-condition assessment.