Masir-AI91

Masir-AI91: A Real-Time Traffic Sign and Obstacle Dataset for Embedded Autonomous Driving Systems Masir-AI91 is a curated image dataset specifically designed for training and evaluating real-time traffic sign recognition and obstacle detection models on embedded platforms. It consists of 91 distinct classes, including regulatory signs, directional cues, and road obstacles, organized into separate train, validation, and test directories. This dataset was used to develop and test an end-to-end autonomous vehicle system based on Raspberry Pi 5, and has been employed in comparative benchmarks of CNN architectures for lightweight deployment. The name Masir (from Arabic مسير) refers to movement, guidance, and path—symbolizing the intelligent navigation capabilities enabled by AI systems. Structure: train/ → 91 class folders with labeled images for model training valid/ → validation data, same class structure test/ → test samples, unseen during training Applications: Lightweight CNN training Benchmarking on edge devices Autonomous navigation and embedded AI research License: CC BY 4.0DOI: 10.5281/zenodo.15347241