
This dataset provides the official test split for the Moroccan License Plate Detection task, supporting the paper "Robust Real-Time Recognition of Multilingual Moroccan License Plates for CPU-Based Edge Devices."
It contains 261 full-vehicle images with corresponding annotations in Pascal VOC (XML) format. The single object class is (license_plate).
This benchmark was created to facilitate reproducible research. The splits were programmatically generated using perceptual hashing to ensure no near-duplicate vehicles exist between the private training sets and this public test set, guaranteeing a leak-free evaluation.