Natural Scene Traffic Sign and Panel Guide Arabic-Latin Text Recognition : NaSTSArLaTR

The presented dataset encompasses a diverse collection of road images captured under a multitude of environmental conditions, specifically sourced from Tunisian highways. Comprising textual annotations in two languages, this dataset is tailored to facilitate research and development in the domain of scene understanding, language processing, and bilingual context analysis. The collection includes 2006 word pictures with Latin and Arabic text occurrences that were taken from 3000 road scene images.  The dataset's versatility enables investigations into the robustness of language models and image recognition systems across different linguistic and environmental contexts.