Moroccan Traffic Sign Recognition

The Moroccan Traffic Sign Recognition (MTSR) dataset is a novel and region-specific benchmark developed to advance both conventional Traffic Sign Recognition (TSR) tasks and Explainable Artificial Intelligence (XAI) research. Comprising 2,111 annotated images across 20 traffic sign categories, the dataset captures diverse real-world conditions from various Moroccan cities and regions. Images were collected under varying lighting, weather, and traffic conditions, and include different angles and occlusions to enhance model generalizability.Beyond standard bounding box annotations (in PASCAL VOC XML format), the dataset provides enriched metadata attributes critical for XAI applications\u2014such as sign color, shape, lighting type, occlusion, environment context (urban\/rural), and camera viewpoint. This enables fine-grained model interpretability using techniques like saliency maps, Score-CAM, LIME, and SHAP. Additionally, the dataset includes a dedicated directory structure for original and preprocessed images, annotations tailored for XAI tools, Score-CAM visualizations, and a bilingual legend of sign classes (Arabic\/English).The MTSR dataset is intended for non-commercial research and academic use, and must be cited in any resulting publication. It provides a valuable resource for developing robust, transparent, and region-aware traffic sign recognition models.