FCDD: A High-Resolution Unstructured Environment Dataset with Multiple Sand Roads

The Floripa Coast Driving Dataset (FCDD) is a 4K-resolution semantic segmentation dataset designed for autonomous driving and unstructured environment perception. It covers a 12.2 km coastal route featuring asphalt, dirt roads, and multiple sand terrains (soft sand, hard sand, and wet sand), captured under diverse weather and lighting conditions.

FCDD comprises 500 pixel-level annotated images (400 for training / 100 for validation) across 17 semantic classes, including road surfaces, markings, cat's eyes, cracks, potholes, vehicles, people, animals, and obstacles. It supports research on terrain classification, road distress detection, and coastal navigation, and enables studies on texture recognition, multi-surface segmentation, and the impact of image resolution.

See the original article: "FCDD: A High-Resolution Unstructured Environment Dataset with Multiple Sand Roads." at IEEE Access.