@Bench: Benchmarking Vision-Language Models for Human-centered Assistive Technology

TL;DR

This paper introduces @Bench, a comprehensive benchmark for evaluating vision-language models in assistive technology for visually impaired people, along with a new multi-task model that improves assistance capabilities.

Contribution

The paper presents a novel benchmark (@Bench) for human-centered assistive tasks and a new multi-task model (@Model) that addresses multiple vision-language tasks simultaneously.

Findings

The benchmark covers five key assistive tasks.

The proposed model outperforms existing methods across tasks.

Experiments demonstrate the model's effectiveness and generalizability.

Abstract

As Vision-Language Models (VLMs) advance, human-centered Assistive Technologies (ATs) for helping People with Visual Impairments (PVIs) are evolving into generalists, capable of performing multiple tasks simultaneously. However, benchmarking VLMs for ATs remains under-explored. To bridge this gap, we first create a novel AT benchmark (@Bench). Guided by a pre-design user study with PVIs, our benchmark includes the five most crucial vision-language tasks: Panoptic Segmentation, Depth Estimation, Optical Character Recognition (OCR), Image Captioning, and Visual Question Answering (VQA). Besides, we propose a novel AT model (@Model) that addresses all tasks simultaneously and can be expanded to more assistive functions for helping PVIs. Our framework exhibits outstanding performance across tasks by integrating multi-modal information, and it offers PVIs a more comprehensive assistance. Extensive experiments prove the effectiveness and generalizability of our framework.