AfriAya — vision-language dataset for African representation in AI
AfriAya, built by Ugandan engineers, adds African cultural context to vision-language models. The dataset spans 13 languages and integrates with Cohere's Aya model to improve AI recognition of African people, places, and traditions.
• Addresses systematic underrepresentation of Africa in existing training data
• Covers diverse regions, languages, and cultural contexts across the continent
• Community-driven effort expanding under Cohere Labs' inclusion initiatives