Benchmarking Vision-Language Models (VLM) for Visual Impairment Assistance in Pakistan using real-world local environments. This project evaluates VLMs like LLaVA-13B and introduces a human-centered, VI-oriented description format focused on navigation, spatial awareness, safety, and accessibility.
benchmarking benchmark accessibility benchmark-framework vlm visual-impairments pakistandataset ai-for-social-good llava assistive-ai vision-language-models
-
Updated
Jan 28, 2026 - Python