VisAssist: A Visually Impaired-Captured Video Question Answering Benchmark for Assistive Systems
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (6) · 4149-4157
DOI: 10.1609/aaai.v40i6.42410 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Qi Gao und 5 weiteren Autor:innen.
Abstract (Originalsprache)
We present VisAssist, the first large-scale video question-answering dataset with 13,413 real-world videos captured by visually impaired users, addressing a critical gap in assistive vision research. Unlike existing benchmarks relying on third-person footage, VisAssist provides authentic first-person perspectives that uniquely capture challenges in blind photography—including unconventional framing, motion artifacts, and frequent information omission. Benchmark evaluations of SOTA multimodal models reveal systematic limitations: severe deficiencies in spatial reasoning when processing dynamic first-person viewpoints, an inability to distinguish missing information from poor capture quality leading to hazardous hallucinations, and fragile text understanding especially for non-Latin scripts under suboptimal conditions. This work establishes a vital real-world benchmark and underscores the need for specialized architectures in visual assistance systems.
Zitieren
APA
Gao, Q., Li, H., Zhou, Y., Zhou, M., Chen, J., & Chai, X. (2026). VisAssist: A Visually Impaired-Captured Video Question Answering Benchmark for Assistive Systems. Proceedings of the AAAI Conference on Artificial Intelligence, 40(6), 4149-4157. https://doi.org/10.1609/aaai.v40i6.42410
Vancouver
Gao Q, Li H, Zhou Y, Zhou M, Chen J, Chai X. VisAssist: A Visually Impaired-Captured Video Question Answering Benchmark for Assistive Systems. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(6):4149-4157. doi:10.1609/aaai.v40i6.42410
BibTeX
@article{Gao2026,
author = {Gao, Qi and Li, Heng and Zhou, Yixin and Zhou, Meixuan and Chen, Jieqiong and Chai, Xinyu},
title = {VisAssist: A Visually Impaired-Captured Video Question Answering Benchmark for Assistive Systems},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {6},
pages = {4149--4157},
doi = {10.1609/aaai.v40i6.42410}
}
EUVIMED