VIABench

This repository contains the data for VIABench: A Comprehensive Video Benchmark Collected from Blind Individuals for Visual Impairment Assistance.

VIABench is a comprehensive egocentric video benchmark for evaluating multimodal large language models in real-world visual impairment assistance scenarios. Collected from videos recorded or shared by blind individuals, VIABench contains 761 videos, 46.9 hours of footage, and 14,526 manually curated annotations across three core tasks.

Tasks

• Proactive Reminder: Evaluates whether a model can understand an ongoing video stream, anticipate navigation-critical events, and provide timely verbal reminders before they occur.
• Visual Question Answering (VQA): Evaluates whether a model can answer user-posed questions about the environment or objects within a video.
• Vision-Guided Interaction: Evaluates context-aware reasoning and guidance for accomplishing intentional interactions between a user and the environment.