Moroccan Monuments Dataset

The Moroccan Monuments Dataset (MMD) is a large-scale, high-quality benchmark designed for semantic segmentation and visual analysis of Moroccan architectural heritage. It comprises 2,686 RGB images collected from historical cities including Fez, Rabat, Marrakech, Meknes, and Tetouan, representing diverse monument types such as mosques, madrasas, royal gates, and mausoleums. Each image is manually annotated at the pixel level into five semantic classes: roof, door, window, wall, and background. The dataset provides a culturally contextualized foundation for training and evaluating computer vision models in digital preservation, restoration planning, and AR\/VR heritage applications. MMD bridges the gap between general-purpose datasets and domain-specific heritage imagery, enabling research in semantic segmentation, cultural informatics, and architectural scene understanding.