ArYSL: Arabic Yemeni sign language dataset.

Area of Science:

• Computer Science
• Artificial Intelligence
• Linguistics

Background:

• Arabic Sign Language (ARSL) recognition faces challenges due to limited datasets, especially for real-world scenarios.
• Existing datasets often lack the diversity and detail needed for robust ARSL translation systems.

Purpose of the Study:

• To introduce the ArYSL Version 2 dataset, an expanded resource for Arabic Sign Language translation.
• To provide a comprehensive dataset that addresses the limitations of previous ARSL resources.

Main Methods:

• Collected 35,900 labeled RGB images of 32 Arabic sign classes from 35 diverse participants.
• Developed a dictionary of 357 Arabic words with synonyms, dialectal variations, and common misspellings.
• Designed the dataset for deep learning and fuzzy logic applications in image-to-text and text-to-image translation.

Main Results:

• The ArYSL Version 2 dataset is a large, dynamic, word-based, and fully-labeled collection.
• The dataset includes detailed Arabic data dictionaries, enhancing its utility for translation.
• The resource is made freely available to the research community.

Conclusions:

• ArYSL Version 2 significantly advances the availability of resources for Arabic Sign Language research.
• The dataset is suitable for developing advanced bidirectional translation systems using AI.
• This contribution aims to foster further research and development in ARSL recognition and translation.