GuideDog dataset aids blind and low-vision navigation with egocentric multimodal data

Researchers have introduced GuideDog, a new dataset designed to aid the development of multimodal large language models (MLLMs) for blind and low-vision (BLV) individuals. The dataset comprises 22,000 image-description pairs from real-world pedestrian scenes across 46 countries, utilizing a human-AI pipeline for more scalable annotation. Additionally, GuideDogQA, an 818-sample benchmark, evaluates object recognition and depth perception, with current MLLMs showing limitations in these areas. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT This dataset could accelerate the development of assistive navigation tools for the visually impaired by providing much-needed real-world data.

RANK_REASON The cluster describes a new academic paper introducing a dataset and benchmark.

Read on arXiv cs.CV →

• GuideDog
• MLLMs
• GuideDogQA
• arXiv

• paper
• product